Item Response Theory Modeling and Categorical Regression Analyses of the Five-Factor Model Rating Form: A Study on Italian Community-Dwelling Adolescent Participants and Adult Participants

Abstract

To extend the evidence on the reliability and construct validity of the Five-Factor Model Rating Form (FFMRF) in its self-report version, two independent samples of Italian participants, which were composed of 510 adolescent high school students and 457 community-dwelling adults, respectively, were administered the FFMRF in its Italian translation. Adolescent participants were also administered the Italian translation of the Borderline Personality Features Scale for Children–11 (BPFSC-11), whereas adult participants were administered the Italian translation of the Triarchic Psychopathy Measure (TriPM). Cronbach α values were consistent with previous findings; in both samples, average interitem r values indicated acceptable internal consistency for all FFMRF scales. A multidimensional graded item response theory model indicated that the majority of FFMRF items had adequate discrimination parameters; information indices supported the reliability of the FFMRF scales. Both categorical (i.e., item-level) and scale-level regression analyses suggested that the FFMRF scores may predict a nonnegligible amount of variance in the BPFSC-11 total score in adolescent participants, and in the TriPM scale scores in adult participants.

Keywords

Five-Factor Model Rating Form item response theory reliability validity adults adolescents

The Five-Factor Model of personality (FFM; McCrae & Costa, 2008) originates from the lexical tradition and comprises five bipolar domains that have been labeled Extraversion (E; or Surgency; vs. Introversion), Agreeableness (A; vs. Antagonism), Conscientiousness (C; vs. Disinhibition), Neuroticism (N; vs. Emotional Stability), and Openness to Experience (O; or Intellect; vs. Closedness to experience). Each of these five broad domains has been differentiated into six more specific facets by Costa and McCrae (1995) on the basis of their development of research with the NEO Personality Inventory–Revised (NEO PI-R). For example, Costa and McCrae suggest that the domain of extraversion (vs. introversion) can be differentiated into the more specific facets of warmth, gregariousness, assertiveness, activity, excitement-seeking, and positive emotionality. Of course, there are certainly other ways to divide the five domains into their component parts (e.g., DeYoung, Quilty, & Peterson, 2007; Lee & Ashton, 2004; Saucier & Goldberg, 2002) and the specific facets delineated by Costa and McCrae (1995) have received some criticism. However, Costa and McCrae’s (1995) hierarchical model of FFM domains prompted the development of a family of assessment instruments, which in turn had a relevant influence on empirical research based on the FFM.

Indeed, facet-level assessments have shown incremental validity beyond the FFM domains for predicting specific behaviors including grade point average attainment and dating frequency (Paunonen & Ashton, 2001; Paunonen, Haddock, Fosterling, & Keinonen, 2003) and have proved to be useful for differentiating among personality disorder constructs (e.g., Axelrod, Widiger, Trull, & Corbitt, 1997; Reynolds & Clark, 2001). For these reasons, measures yielding scores for the lower order facets, in addition to the higher order domains, are deemed to be particularly useful in many contexts. Perhaps the most widely used facet-level measure of the FFM is the NEO PI-R (Costa & McCrae, 1992). The NEO PI-R is a self-report inventory that was explicitly designed to assess 30 facets and 5 domains of the FFM as they were delineated by Costa and McCrae (1995). The NEO PI-R has long been recognized as the one of the most well-validated measures of the FFM (Briggs, 1992).

Notwithstanding the compelling measurement properties of the NEO-PI-R, an obstacle to its use is that it requires at least 20 to 30 minutes to complete. For this reason, Lynam and Widiger (2001) developed a brief rating form that allowed them to collect facet-level FFM descriptions of personality disorders from expert researchers and clinicians. In doing so, they included the identifying label for each of the NEO PI-R facet scales, as well as a few adjective descriptors to describe each of the poles. This rating form, which serves as a brief measure of the NEO PI-R model, was subsequently termed the Five-Factor Model Rating Form (FFMRF). The FFMRF showed adequate interrater reliability and evinced temporal consistency comparable to other FFM measures (i.e., median 6-month stability for the domains was .54; Samuel & Widiger, 2011).

Starting from these premises, Mullins-Sweatt, Jamerson, Samuel, Olson, and Widiger (2006) used the FFMRF to collect self-report ratings. Through a series of studies, they made minor revisions to the adjective descriptors (e.g., sensitive and responsive were replaced by self-aware for high openness to feelings), developing a 30-item self-report measure of the FFM facets and domains, yielding a single item for each facet and a total score for each FFM domain, which was simply the sum of the corresponding six items. The FFMRF domain scores displayed substantial convergent validity coefficients that were markedly larger than the corresponding discriminant validity coefficients (Mullins-Sweatt et al., 2006; Samuel, Mullins-Sweatt & Widiger, 2013); similar considerations held also for facet-level analyses. The FFMRF scores manifested predictable relationships with personality pathology, supporting the construct validity of the measure (Mullins-Sweatt et al., 2006). Recently, Samuel et al. (2013) reported exploratory structural equation model findings supporting the five-factor structure of the FFMRF in undergraduate students. The FFMRF has since been used in other studies as a self-report measure (e.g., Howell, Dopko, Turowski, & Buro, 2011; Schenk, Ragatz, & Fremouw, 2012; Thomas et al., 2013).

Notwithstanding the promising findings, further studies on the psychometric properties of the FFMRF may be worth carrying out. The FFMRF development did not strictly follow the domain-sampling model (Nunnally & Bernstein, 1994); for instance, each FFMRF items is considered a valid indicator of the corresponding facet, rather than one of the many possible alternative indicators that could be (randomly) sampled from the corresponding domain. Under these circumstances, item response theory (IRT) methodologies may be well suited for a more in-depth examination of the properties of the FFMRF items and their relations with latent traits (e.g., Edelen & Reeve, 2007). Indeed, IRT analyses contrast from classical test theory by focusing on latent properties of items rather than observed scores (Embretson & Reise, 2000). Nevertheless, no study on the psychometric properties of the FFMRF using an IRT approach has been carried out. Moreover, to our knowledge, no independent replication of the original findings on the FFMRF has been conducted, particularly in a cross-cultural (or at least cross-language) perspective. Furthermore, no data on the psychometric properties of the FFMRF in adolescent samples, as well as in samples of adult participants different from undergraduate student sample are currently available. Finally, the FFMRF represents the sole brief measure that permits assessment of not only the 5 domains but also the 30 facets included within the NEO PI-R. The inclusion of the lower order facets is crucial for the utility of the FFMRF (Samuel et al., 2013), as they have been shown to be useful for differentiating constructs (Reynolds & Clark, 2001) and predicting specific behaviors (Paunonen et al., 2003). Indeed, meta-analytic studies showed that facet-level assessment of FFM constructs yielded clinically meaningful profiles, which efficiently discriminated different manifestations of personality pathology (e.g., Samuel & Widiger, 2008). Thus, the FFMRF could be of some interest for clinicians because it provides scores for both FFM facets and domains while being an easy-to-administer (and score) instrument. Notwithstanding these attractive features of the FFMRF, to our knowledge, no study attempted to evaluate the usefulness of FFMRF facet scores as predictors of measures of clinically relevant external constructs.

Against this background, in the present study we were interested in evaluating the presence of significant associations between FFMRF facet (i.e., item-level) scores and selected personality disorder dimensions. We tested hypotheses concerning the clinical usefulness of the FFMRF in a nomological network perspective with respect to different pathological personality dimensions in adolescent participants and in adult participants, respectively. In selecting these personality disorder dimensions, we considered the following aspects: (a) clinical and/or social relevance of the personality pathology dimension even in community dwelling samples (e.g., necessity of early assessment and intervention, relevance of associated problems, such as drug use, self-destructive behavior, or physical aggression, etc.); (b) existence of established FFM trait profiles; and (c) availability of Italian versions of the corresponding measures which were provided with reliability and validity data in community dwelling adolescents and/or adults.

Based on these considerations, we focused our attention on borderline personality disorder and psychopathy dimensions. The assessment of borderline personality pathology in adolescence is important, since personality dysfunction is commonly misdiagnosed or missed completely in the adolescent population, in which emotional dysregulation and externalizing behavior can be explained as developmentally appropriate, depending on its magnitude (Sharp & Bleiberg, 2007). Starting from these considerations, Sharp, Steinberg, Temple, & Newlin (2014) proposed an 11-item version of the Borderline Personality Features Scale for Children (BPFSC; Crick, Murray-Close, & Woods, 2005). Sharp et al., (2014), using a sample of 371 inpatient adolescents, demonstrated similar indices of construct validity as observed for the BPFSC total score with the BPFSC-11 scores and found evidence for good criterion validity. Fossati, Sharp, Borroni, and Somma (2015) showed that the Italian translation of the BPFSC-11 was provided with adequate reliability (Cronbach α = .78), moderate 6-month test–retest stability (e.g., Pearson r value for the test–retest correlation of the BPFSC-11 total score was .50, p < .001), factor structure validity (i.e., root mean square error of approximation [RMSEA] = .04, close fit p > .70), and convergent validity (the bivariate correlation between the BPFSC-11 and the PDQ-4 + BPD scale was .64, p < .001) in a sample of 805 community dwelling adolescents. Unfortunately, no adult version of the BPFSC-11 currently exists.

Research findings have consistently documented associations between psychopathy and a wide range of externalizing behaviors such as crime and aggression (e.g., Gretton, Hare, & Catchpole, 2004), criminal recidivism (e.g., Walters, Knight, Grann, & Dahle, 2008), substance use (e.g., Kennealy, Hicks, & Patrick, 2007), and sexual offending (Caldwell, Ziemke, & Vitacco, 2008). Starting from these perspectives, Patrick, Fowles, and Krueger (2009) proposed a triarchic model of psychopathy. The essence of the triarchic model is that psychopathy encompasses three distinct phenotypic constructs: (a) boldness, which is defined as the nexus of social dominance, emotional resiliency, and venturesomeness; (b) meanness, which is defined as aggressive resource seeking without regard for others (“dysaffliated agency”); and (c) disinhibition, which reflects a general propensity toward problems of impulse control. Patrick (2010) proposed the Triarchic Psychopathy Measure (TriPM) as a self-report instrument, which was explicitly designed to assess the triarchic model of psychopathy. Evidence has emerged for the effectiveness of the TriPM scales as indices of the triarchic model constructs (for a review, see Patrick & Drislane, 2015). Interestingly, Poy, Segarra, Esteller, López, and Moltó (2014) provided evidence that the relationships between TriPM scales and measure of FFM traits were consistent with the conceptualization of psychopathy in terms of FFM constructs in community-dwelling adults. Recently, Sica et al. (2015) provided reliability and validity data of the Italian translation of the TriPM in community dwelling adults.

Thus, in the present study we assessed borderline personality disorder dimension using the BPFSC-11 in community-dwelling adolescents, while relying on the TriPM to assess psychopathy in community dwelling adults. Starting from these considerations we designed the present study with the following aims:

Testing a full-information confirmatory multidimensional IRT model based on the graded response model (Samejima, 1997), in which five factors were specified. IRT models were carried out in both adolescent sample and adult sample.

Evaluating the psychometric properties of the FFMRF in a different cultural context from the United States (i.e., Italy), and in different age groups (i.e., adolescence and adulthood).

Testing if the FFMRF items (i.e., facets) and domain scales, respectively, predict selected dysfunctional personality trait measures consistent with prior research that has used more extensive and established measures of the FFM. In particular, based on existing meta-analyses on the association between FFM facets (Samuel & Widiger, 2008), we expected that the BPFSC-11 total score was significantly predicted by high scores on FFMRF Angry Hostility (+), Impulsiveness (+), Warmth (−), Positive Emotions (−), Trust (−), Compliance (−), Dutifulness (−), Self-discipline (−), and Deliberation (−). Considering the FFM domains that were reported as substantial and consistent (i.e., no significant heterogeneity among studies) features of BPD in Samuel and Widiger’s (2008) meta-analysis, we expected that high score on FFMRF Neuroticism (N) scale, and low scores on FFMRF Agreeableness (A) and Conscientiousness (C) domain scales significantly predicted the BPFSC-11 total score. Similarly, based on Poy et al.’s (2014) findings, who used the NEO-PI to identify the FFM facets (and domains) that significantly predicted a triarchic measure of psychopathy (i.e., the TriPM) in undergraduate students, we expected that the TriPM Boldness scale score was significantly predicted by all FFMRF N facets (apart from Impulsiveness), and all E facets, as well as by Actions (O), Ideas (O), Straightforwardness (A), Modesty (A), and Competence (C); we also expected that TriPM Meanness scale score significantly predicted by low scores on all FFMRF A facets, and possibly also Dutifulness (C) and Discipline (C) facets. We hypothesized that TriPM Disinhibition scale score was significantly predicted by FFMRF C facets, A facets, and N facet scores. Finally, in our adult sample we tried to evaluate for the first time the relationships between the TriPM total score, as an index of the overall psychopathy level according to the triarchic model, and the FFMRF domain scale scores and facet scores, respectively.

Method

Participants

Adolescent Sample

Five hundred thirteen adolescent high school students attending a public high school in Rome agreed to participate in the study. Since three participants (0.6%) reported more than 10% of missing responses on the FFMRF and BPFSC-11 items, they were dropped from the study. The number of participants with missing values was too small to carry out formal missing value analyses. The final sample was comprised 510 adolescent high school students; 308 participants (60.4%) were female and 202 (39.6%) were male; the mean age was 16.31 years, SD = 1.38 years, range = 14 to 19 years. All adolescent participants were unmarried. To participate in the present study, participants were required to speak Italian as their first language in order to avoid cultural and lexical bias in questionnaire responses. None of the participants received an economic incentive to participate in the study.

After obtaining institutional review board approval from the university and the principals of the schools, researchers recruited adolescents from classrooms (data were collected in autumn 2013 to spring 2014). Written informed parent consent and adolescent assent were obtained prior to study participation. Although socioeconomic status was not directly controlled for in this study, it should be emphasized that participants were selected from public high schools. Data from the National Institute of Statistics of Italy (retrieved from http://www.istat.it/it/archivio/17290) showed that 93.1% of Italian adolescents were high school students during 2012/2013, thus suggesting that adolescents attending public high schools who took part in the present set of studies are likely to be representative of the Italian adolescent population. Participants received the Italian translations of the instruments. The questionnaires were administered in random order and anonymously during class time in school by graduate research assistants when teachers were not present in the classrooms.

Adult Sample

Four hundred and sixty-six community-dwelling adults originally agreed to participate in this study; none of the participants received an economic incentive to take part in the study. Participants were community-dwelling adults living in Rome metropolitan area who responded to advertisements requesting potential volunteers for psychological studies that were placed at the LUMSA university campus and on the web during autumn 2013 to spring 2014. All participants gave their written consent to participate in the study after it had been explained to them. Since 10 participants (2.1%) yielded more than 10% of missing responses on the FFMRF and TriPM items, they were excluded from the study. Participants who reported missing observations did not differ significantly from participants who reported complete questionnaires on participants’ gender, χ²(1) = 2.14, p > .10, ϕ = .07; age, t(464) = 1.33, p > .10, d = 0.12; civil status, χ²(3) = 1.74, p > .10, Cramer V = .06; and school degree, χ²(2) = 3.50, p > .10, ϕ = .09. Missing value analyses bases on expectation maximization algorithm showed that missing values were completely at random considering for FFMRF items, Little χ²(86) = 86.12, p > .40, and scales, Little χ²(8) = 4.22, p > .80; similar findings were observed for the missing value pattern on the TriPM.

The final version of this sample comprised 457 Italian community-dwelling adults who were living in Rome metropolitan area; 260 participants (56.9%) were female and 197 (43.1%) were male; participants’ mean age was 38.33 years, SD = 12.39 years. Two hundred fifty-nine participants were unmarried (56.7%), 150 (32.8%) were married, 34 (7.4%) were divorced, and 4 (0.9%) were widow/widower; 10 participants (2.2%) did not report their civil status. Thirty-five participants (7.7%) had junior high school degree, 215 (47.0%) had high school degree, and 202 (44.2%) had university degree (i.e., BSc or higher); five participants (1.1%) refused to disclose their school degree. All participants were active community members (i.e., none of the participants was retired) when the study was carried out. Participants received the Italian translations of the instruments. The questionnaires were administered in random order and anonymously in individual sessions by graduate research assistants.

Measures

Five Factor Model Rating Form (Mullin-Sweatt et al.,2006)

The FFMRF consists of 30 self-report items representing each of the 30 facets of the FFM organized with respect to the five domains, that is, Neuroticism (N), Extraversion (E), Openness to Experience (O), Agreeableness (A), and Conscientiousness (C). Each item is rated on a 1 (extremely low) to 5 (extremely high) scale. Item scores are summed to yield the total score of the corresponding FFMRF scale dimension. Data have supported the reliability and validity of the FFMRF (Mullins-Sweatt et al., 2006; Samuel et al., 2013) in both clinical and nonclinical adult samples.

Borderline Personality Features Scale-11 (Sharp et al.,2014)

The BPFSC-11 was developed through IRT analyses from the original 24-item BPFSC (Crick et al., 2005) and consists of 11 items measuring borderline personality features in childhood (for ages 9 and older, including adolescents). These items assess how participants feel about themselves and other people, and are rated on a 5-point Likert-type scale ranging from not true at all to always true. The BPFSC-11 yields a total score measuring the overall level of borderline characteristics; the higher the BPFSC-11 total score, the greater the intensity of BPD features. The BPFSC-11 has shown adequate reliability (Cronbach α = .85; Sharp et al., 2014); in terms of construct validity, the BPFSC-11 showed adequate sensitivity (.74) and specificity (.71) with respect to DSM-IV BPD diagnosis, as well as significant correlations with measures of self-harm and emotion dysregulation, in a sample of adolescent inpatients (Sharp et al., 2014). The BPFSC-11 showed adequate psychometric properties also in its Italian translation (Fossati et al., 2015).

Triarchic Psychopathy Measure (Patrick, 2010)

The TriPM is a 58-item self-report inventory that includes three scales for indexing the phenotypic components of psychopathy specified by the triarchic model of psychopathy: boldness, meanness, and disinhibition. Participants respond to each item on a 4-point Likert-type scale (3 = true, 2 = mostly true, 1 = mostly false, 0 = false). The Boldness scale comprises 19 items that index tendencies toward social poise and effectiveness, emotional resiliency, and venturesomeness. The Disinhibition and Meanness scales (20 and 19 items, respectively) index broad disinhibition and callous-aggression factors, respectively. Scores on the three scales are summed to yield an overall triarchic psychopathy score (Patrick, 2010; Patrick & Drislane, 2015). Recent published research provides support for the validity of the TriPM as a measure of psychopathic features (e.g., Drislane, Patrick, & Arsal, 2014; Sellbom & Phillips, 2013), also in its Italian translation (Sica et al., 2015).

Measure Translation Procedures

Equivalence with the original meaning of the items was the guiding principle in the translation process (Denissen, Geenen, van Aken, Gosling, & Potter, 2008). First, the FFMRF was independently translated into Italian by one of the authors (AF), and by two other clinical psychologists who were fluent in English. After reaching a consensus, we had an English mother-tongue professional translator translate the Italian version back into English, and this English back-translation (Cha, Kim, & Erlen, 2007; Geisinger, 1994; Van de Vijver & Hambleton, 1996) was sent to the author of the FFMRF. If the latest version differed from the English original, the translators came to an agreement on the definitive Italian translation. The authors followed the same procedure of translation concerning the BPFSC-11 and TriPM, although co-translators were different. In the present study, the same Italian translation of the TriPM that was previously used in Sica et al.’s (2015) study was administered.

Data Analyses

A full-information confirmatory multidimensional IRT model in which five factors were specified was used to evaluate the five-factor structure of the FFMRF. Multidimensional IRT models were selected in this study as the main method for different reasons. First, one advantage of IRT models over linear factor analytic methods is that information from examinee response patterns is analyzed as opposed to the more limited information from correlation matrices (Lane & Stone, 2006). Second, nonlinear models such as IRT models may better reflect the relationship between item performance and the latent trait (Hattie, 1985). The graded response model (Samejima, 1997) was used as the item response model; indeed, it has often been found useful for questionnaire data collected using Likert-type scales (e.g., Samuel, Simms, Clark, Livesley, & Widiger, 2010). The IRT approach was used to assess item discrimination (i.e., validity) and thresholds (i.e., difficulty) for each FFMRF domains. Goodness of fit of the IRT multidimensional model was evaluated using the M₂* statistics and its associated RMSEA value (Cai & Hansen, 2013; Cai, Maydeu-Olivares, Coffman, & Thissen, 2006). The standardized local dependence (LD) chi-square statistic (based on the LD statistic proposed by Chen & Thissen, 1997) and Q₃ (Yen, 1984) statistic were used to further evaluate the goodness of fit of the IRT models. Standardized LD values greater of 10 or greater are considered noteworthy (Chen & Thissen, 1997; for an application, see Sharp et al., 2014). Q₃ is the Pearson product-moment correlation of a set of residuals from the IRT model; Q₃ absolute values ≥.20 are usually considered indicative of departures from the local independence assumption (Chen & Thissen, 1997).

To increase the comparability of our findings with previous results on the FFMRF scale reliability, we computed Cronbach α coefficient as a measure of internal consistency reliability of FFMRF scales. However, the usefulness of Cronbach α coefficient in assessing the measure reliability has been called into question, particularly for short scales (Clark & Watson, 1995). Interestingly, IRT provides a different approach to measurement reliability that is based on test information, rather than on test length and average interitem correlation (Thissen, 1990). IRT operationalizes standard error of measurement as one divided by the square root of test information; from this perspective test reliability is simply one minus error variance (i.e., the square of standard error of measurement; Thissen, 1990). The IRT reliability coefficient ranges from 0.00 in the case of a completely unreliable test to 1.00 in the case of a perfectly reliable test. Since one major achievement of IRT models is greater precision in describing how much information a measure provides about an underlying latent trait (Markon, 2013a), the test information curves of the FFMRF domain scales were provided as a visual depiction of where along the trait continuum the FFMRF domain scales were most discriminating (Reise & Waller, 2002). The test information function specifies how much psychometric information a measure provides about a specific trait level, and it supplements and extends traditional notions of reliability (Markon, 2013a). IRT approximation of reliability for the FFMRF domain scores in the −3, 3 range was calculated as one minus the measurement variance (Thissen, 1990).

Since a burgeoning literature is providing new advances on quantifying information—and consequently reliability—in IRT analysis, in the present study, we tried to quantify the information utility, which reflects the degree to which a measure increases the posterior probability of trait estimates that are in fact most probable, relative to their prior probability (Markon, 2013a). To this purpose, the following information utilities indices (Markon, 2013a) were computed for each FFMRF domain scale: (a) expected information utility (EIU); (b) the information utility lower bound and upper bound; and (c) normalized minimum reduction in utility (NMRU). EIU is similar to test reliability in that it provides an index of overall measurement quality in a sample or group of individuals. The larger the EIU value the larger the amount of explained information; EIU values greater than 1.00 indicates that the test explains proportionally more information than the reference prior (in our analyses, Jeffreys prior was used to compute reference prior; Clarke & Barron, 1994; Ghosh, Mergel, & Liu, 2011; Jeffreys, 1946). The NMRU, in contrast, is invariant across samples, and provides estimates of how well the test is measuring globally, being both sample- and response-independent (Markon, 2013a). NMRU values are bounded between 1.00 and 0.00; tests with large values of the NMRU have large “potential,” in the sense that they provide large amounts of information—or conversely, reduce uncertainty (Markon, 2013a). EIU and NMRU were computed by fitting full information maximum likelihood IRT models separately to each FFMRF scale because currently the algorithm for computing these information indices does not allow to manage multidimensional data (Markon, 2013a).

Categorical regression analysis (CATREG) is a nonparametric method to perform multiple regression when data are categorical (van der Kooij, Meulman, & Heiser, 2006). In the present study, CATREG was used to evaluate if FFMRF items significantly predicted the BPFSC-11 total score, and the TriPM scales and total score, in the adolescent sample, and in the adult sample, respectively. CATREG quantifies categorical variables using optimal scaling, resulting in an optimal linear regression equation (van der Kooij et al., 2006). For selecting a subset of stable predictors, we used the Least Absolute Shrinkage and Selection Operator (LASSO; Tibshirani, 1996), which was incorporated into the CATREG method (van der Kooij et al., 2006). The LASSO combines the improved prediction accuracy of Ridge regression in the presence of multicollinearity (Hoerl & Kennard, 1970a, 1970b) with the better interpretability of subset selection. The shrinkage is accomplished by adding a penalty term to the regression model, penalizing the sum of the squared regression coefficients; the LASSO applies a penalty to the sum of the absolute values of the coefficients. With a penalty value of zero, the model including all predictors (unshrunken) is obtained and with a high penalty value all predictors are shrunken to zero. Penalty values in between zero and some high value result in models with different shrunken coefficients with some of them shrunken to zero. Thus, different values of the penalty result in different models. The optimal value of the penalty was determined using the .632 bootstrap (Efron, 1983; for details of using the .632 bootstrap procedure with CATREG, see van der Kooij & Meulman, 2004), which is essentially a smoothed version of leave-one-out cross-validation to estimate expected prediction error (and its standard error). The .632 bootstrap procedure has to be applied to each LASSO model (starting with a penalty value of zero and increasing the penalty value in small steps until all predictors are shrunken to zero). To select the optimal LASSO model usually the one-standard-error rule is applied: select the most parsimonious LASSO model within one standard error of the LASSO model with minimum expected prediction error.

The R (R Core Team, 2015) package “mirt” (Chalmers, 2012) was used to perform multidimensional IRT model; information utility indices were computed using the R package “infutil” (Markon, 2013b). Regression analyses were computed using SPSS, version 22.

Results

The descriptive statistics and Cronbach α values for the FFMRF domain scale scores in adolescent high school students and community-dwelling adults, respectively, are listed in Table 1. Among adolescent high school students, average interitem r values were .20, .30, .18, .24, and .28 for FFMRF N, E, O, A, and C domain scales, respectively; in the community-dwelling adult sample, average interitem r values for FFMRF N, E, O, A, and C domain scales were .23, .32, .26, .25, and .43, respectively.

Table 1.

Descriptive Statistics, Correlation (Pearson r) Coefficients, and Cronbach Alpha Values (Main Diagonal) for the Five-Factor Model Rating Form Domain Scales in Adolescent High School Students (N = 510) and Community-Dwelling Adults (N = 457).

	Adolescent high school students (N = 510)
	M	SD	1	2	3	4	5
1. Neuroticism (FFMRF)	16.72	3.89	.61
2. Extraversion (FFMRF)	22.28	3.73	−.28	.71
3. Openness to Experience (FFMRF)	21.33	3.18	.08	.34	.56
4. Agreeableness (FFMRF)	21.52	3.47	.01	.25	.24	.66
5. Conscientiousness (FFMRF)	21.77	3.44	−.10	.22	.14	.26	.70
	Community-dwelling adults (N = 457)
	M	SD	1	2	3	4	5
1. Neuroticism (FFMRF)	16.47	3.70	.64
2. Extraversion (FFMRF)	21.52	3.68	−.29	.74
3. Openness to Experience (FFMRF)	20.34	3.51	−.02	.39	.67
4. Agreeableness (FFMRF)	21.79	3.17	−.03	.21	.22	.66
5. Conscientiousness (FFMRF)	22.09	3.78	−.20	.28	.08	.21	.81

Note. FFMRF = Five-Factor Model Rating Form. The nominal significance level (i.e., p < .05) was corrected according to the Bonferroni procedure and set at p < .002; bold highlights significant r coefficients.

Full-Information Confirmatory Multidimensional Item Response Theory Results

Confirmatory factor analysis, dimensionality analysis, and exploratory structural equation model results of the Italian translation of the FFMRF in the adolescent sample and in the adult sample, respectively, is provided in the online Supplemental Materials (available online at http://journals.sagepub.com/doi/suppl/10.1177/1073191115621789).

Full-information IRT was used to evaluate the fit of the confirmatory multidimensional IRT model in which five factors were specified according to the a priori item-to-scale assignment of the 30 items composing the FFMRF. In IRT local dependence (LD) occurs when items are more strongly correlated than can be accounted for by the general factor (e.g., N; Chen & Thissen, 1997). In our adolescent sample, all standardized LD chi-square values (Chen & Thissen, 1997) were lower than 10.00, with the exception of a standardized LD value of 12.17 for Item 28 (“Achievement; workaholic, ambitious vs. aimless, desultory”) and Item 29 (“Self-Discipline; dogged, devoted vs. hedonistic, negligent”); however, the corresponding Q₃ value (i.e., the Pearson product-moment correlation of a set of residuals from the IRT model; Chen & Thissen, 1997) was −.07, well below the cutoff value of .20 (in absolute value). Similar findings were observed in the adult sample; a standardized LD value greater than 10.00 (i.e., 14.61) was observed only for Item 20 (“Straightforwardness; confiding, honest vs. cunning, manipulative, deceptive”) and Item 21 (“Altruism; sacrificial, giving vs. stingy, selfish, greedy, exploitative”), but the corresponding Q₃ value was only −.04.

Slope and thresholds parameters of the FFMRF items, information function, IRT approximation of reliability (Thissen, 1990), and information utility indices (Markon, 2013a) for the FFMRF domains in the adolescent sample and in the adult sample are shown in Table 2 and in Table 3, respectively. FFMRF item slope and threshold parameters and associated standard errors were estimated simultaneously based on five-factor confirmatory IRT model. In IRT, the slope parameters correspond to the indicator’s ability to discriminate between individuals and are also referred to as the discrimination parameter. Slope parameters are algebraically related to factor loadings, and if they are greater than .90, they are considered substantial (e.g., Forero, Vilagut, Adroher, Alonso, & ESEMeD/MHEDEA Investigators, 2013). Thresholds correspond to the level of the latent trait that is required for an individual to endorse a given response. Thresholds are often analogized as the item’s difficulty, but within personality and assessment, it might more accurately be referred to as extremity or location. In both samples, all discrimination parameters were statistically significant (all ps < .001). In the adolescent sample, 24 FFMRF items (80.0%) showed discrimination parameter values equal to, or greater than 0.90, whereas only one item (3.3%) showed discrimination parameter values lower than 0.50. Similar findings were observed in the adult sample where 26 FFMRF items (86.7%) presented discrimination parameter values of 0.90 or greater, whereas none of the discrimination parameter value was lower than 0.50.

Table 2.

Five-Factor Model Rating Form Scales: Slope, Threshold Parameter Estimates, Standard Errors, Information Function, IRT Approximation of Reliability, and Information Utility Indices in the Adolescents High School Student Sample (N = 510).

	a (SE)	b₁ (SE)	b₂ (SE)	b₃ (SE)	b₄ (SE)
Neuroticism
1. Anxiousness	1.71 (0.21)	3.30 (0.25)	1.52 (0.16)	−0.45 (0.13)	−2.94 (0.23)
2. Angry Hostility	0.50 (0.12)	1.96 (0.14)	0.14 (0.09)	−1.35 (0.11)	−3.19 (0.22)
3. Depressiveness	1.43 (0.19)	0.86 (0.13)	−0.64 (0.12)	−2.18 (0.18)	−3.97 (0.30)
4. Self-consciousness	1.10 (0.15)	2.70 (0.19)	1.12 (0.12)	−0.49 (0.11)	−2.22 (0.16)
5. Impulsivity	0.36 (0.12)	2.07 (0.14)	0.56 (0.09)	−0.78 (0.10)	−2.21 (0.15)
6. Vulnerability	1.47 (0.18)	2.02 (0.17)	−0.08 (0.12)	−1.89 (0.16)	−3.64 (0.26)
Total information/IRT reliability		12.32 (76.95%)/.92
EIU (IU upper and lower bounds)		1.45 (1.43-3.11)
NMRU	.46
Extraversion
7. Gregariousness	1.06 (0.14)	5.08 (0.47)	2.79 (0.19)	1.13 (0.12)	−0.82 (0.11)
8. Assertiveness	1.92 (0.21)	5.15 (0.42)	3.34 (0.26)	1.04 (0.15)	−1.37 (0.16)
9. Activity	1.01 (0.14)	2.95 (0.20)	1.22 (0.12)	−0.98 (0.11)	−2.92 (0.20)
10. Excitement-Seeking	1.58 (0.18)	5.10 (0.42)	3.18 (0.23)	0.83 (0.13)	−1.21 (0.14)
11. Positive Emotions	1.11 (0.15)	3.69 (0.26)	1.90 (0.15)	−0.09 (0.11)	−1.77 (0.14)
12. Gregariousness	1.85 (0.21)	6.26 (0.59)	4.38 (0.33)	1.97 (0.18)	−0.58 (0.13)
Total information/IRT reliability		15.27 (69.51%)/.93
EIU (IU upper and lower bounds)		1.43 (1.41-2.77)
NMRU	.51
Openness to Experience
13. Fantasy	0.93 (0.16)	3.82 (0.29)	2.16 (0.16)	0.67 (0.11)	−0.86 (0.11)
14. Aesthetics	0.88 (0.16)	3.66 (0.27)	2.12 (0.15)	0.08 (0.10)	−1.99 (0.15)
15. Feelings	0.66 (0.14)	5.26 (0.58)	2.62 (0.18)	0.51 (0.10)	−1.58 (0.13)
16. Actions	1.37 (0.19)	3.72 (0.28)	1.83 (0.16)	−1.25 (0.14)	−3.37 (0.24)
17. Ideas	1.73 (0.25)	4.82 (0.42)	2.83 (0.24)	0.52 (0.13)	−2.31 (0.21)
18. Values	0.73 (0.14)	3.80 (0.29)	1.93 (0.14)	−0.04 (0.10)	−1.84 (0.14)
Total information/IRT reliability		10.41 (64.84%)/.90
EIU (IU upper and lower bounds)		1.48 (1.47-3.14)
NMRU	.47
Agreeableness
19. Trust	0.90 (0.14)	2.88 (0.20)	0.97 (0.11)	−0.88 (0.11)	−2.58 (0.17)
20. Straightforwardness	1.08 (0.15)	4.53 (0.38)	2.45 (0.17)	0.44 (0.11)	−1.71 (0.14)
21. Altruism	1.71 (0.21)	5.12 (0.44)	3.42 (0.26)	−1.12 (0.14)	−1.59 (0.16)
22. Compliance	1.39 (0.17)	3.87 (0.29)	1.94 (0.16)	−0.28 (0.11)	−2.95 (0.21)
23. Modesty	1.11 (0.15)	4.12 (0.32)	2.78 (0.19)	−0.48 (0.11)	−1.64 (0.14)
24. Tender-mindedness	1.55 (0.20)	4.96 (0.42)	2.94 (0.22)	0.70 (0.12)	−1.32 (0.14)
Total information/IRT reliability		13.21 (66.8%)/.92
EIU (IU upper and lower bounds)		1.41 (1.40-2.83)
NMRU	.49
Conscientiousness
25. Competence	1.65 (0.18)	4.91 (0.39)	3.22 (0.23)	0.52 (0.13)	−2.60 (0.20)
26. Order	1.08 (0.14)	3.15 (0.21)	1.61 (0.13)	−0.13 (0.11)	−2.08 (0.15)
27. Dutifulness	1.54 (0.18)	5.21 (0.45)	3.85 (0.28)	0.80 (0.13)	−2.15 (0.17)
28. Achievement Striving	1.45 (0.17)	6.00 (0.63)	3.51 (0.25)	0.81 (0.13)	−1.52 (0.14)
29. Self-Discipline	1.84 (0.22)	6.95 (0.81)	3.87 (0.30)	0.23 (0.13)	−2.69 (0.22)
30. Deliberation	0.86 (0.13)	3.90 (0.29)	2.19 (0.15)	0.17 (0.10)	−1.50 (0.12)
Total information/IRT reliability		15.88 (67.42%)/.94
EIU (IU upper and lower bounds)		1.47 (1.46-2.78)
NMRU	.52

Note. a = slope parameters; b = threshold parameters; SE = standard error; Total information = information function in the −3, 3 range; IRT reliability = item response theory approximation of reliability; EIU = Expected Information Utility; NMRU = Normalized Minimum Reduction in Uncertainty; IU = Information Utility. Bold highlights significant (i.e., p < .001) slope parameters.

Table 3.

Five-Factor Model Rating Form scales: Slope, Threshold Parameter Estimates, Standard Errors, Information Function, IRT Approximation of Reliability, and Information Utility Indices in the Community Dwelling Adult Sample (N = 457).

	a (SE)	b₁ (SE)	b₂ (SE)	b₃ (SE)	b₄ (SE)
Neuroticism
1. Anxiousness	1.39 (0.23)	2.85 (0.28)	1.28 (0.19)	−0.71 (0.16)	−3.18 (0.30)
2. Angry Hostility	1.14 (0.20)	1.79 (0.19)	−0.28 (0.15)	2.00 (0.21)	−4.86 (0.72)
3. Depressiveness	1.73 (0.29)	1.63 (0.23)	−0.22 (0.17)	−2.24 (0.28)	−4.96 (0.64)
4. Self-consciousness	0.88 (0.17)	2.42 (0.23)	0.84 (0.15)	−1.03 (0.16)	−3.17 (0.31)
5. Impulsivity	0.57 (0.15)	2.29 (0.22)	0.57 (0.14)	−0.98 (0.15)	−2.67 (0.26)
6. Vulnerability	1.15 (0.19)	3.07 (0.29)	0.49 (0.15)	−1.63 (0.19)	−3.99 (0.43)
Total information/IRT reliability		13.58 (70.46%)/.93
EIU (IU upper and lower bounds)		1.47 (1.46-3.01)
NMRU	.49
Extraversion
7. Gregariousness	2.11 (0.82)	5.89 (0.75)	3.48 (0.35)	1.35 (0.21)	−1.48 (0.23)
8. Assertiveness	2.47 (0.35)	6.42 (0.79)	3.25 (0.35)	1.08 (0.22)	−1.83 (0.28)
9. Activity	0.97 (0.17)	4.07 (0.50)	1.90 (0.20)	−0.62 (0.15)	−2.65 (0.25)
10. Excitement-Seeking	1.27 (0.19)	4.95 (0.74)	3.17 (0.34)	0.61 (0.16)	−1.64 (0.19)
11. Positive Emotions	0.99 (0.19)	3.55 (0.36)	1.16 (0.16)	−1.02 (0.15)	−2.68 (0.25)
12. Gregariousness	1.72 (0.23)	5.56 (0.69)	3.50 (0.30)	1.27 (0.20)	−1.56 (0.21)
Total information/IRT reliability		19.77 (74.18%)/.95
EIU (IU upper and lower bounds)		1.51 (1.50-2.77)
NMRU	.54
Openness to Experience
13. Fantasy	1.18 (0.18)	3.06 (0.26)	1.35 (0.16)	−0.01 (1.14)	−1.77 (0.18)
14. Aesthetics	1.16 (0.18)	4.44 (0.51)	1.76 (0.19)	−0.32 (0.14)	−2.68 (0.26)
15. Feelings	0.61 (0.15)	5.11 (1.20)	3.17 (0.34)	0.90 (0.15)	−1.51 (0.18)
16. Actions	1.97 (0.29)	4.30 (0.46)	1.68 (0.22)	−1.38 (0.21)	−4.28 (0.47)
17. Ideas	2.73 (0.41)	5.72 (0.80)	2.15 (0.30)	−0.99 (0.23)	−4.29 (0.52)
18. Values	0.93 (0.16)	4.16 (0.48)	1.86 (0.19)	−0.13 (0.13)	−2.23 (0.20)
Total information/IRT reliability		19.76 (78.67%)/.95
EIU (IU upper and lower bounds)		1.58 (1.58-2.99)
NMRU	.53
Agreeableness
19. Trust	0.73 (0.17)	3.50 (0.34)	1.27 (0.15)	−0.65 (0.14)	−2.63 (0.26)
20. Straightforwardness	1.18 (0.21)	6.66 (2.15)	3.77 (0.39)	1.08 (0.80)	−1.49 (0.18)
21. Altruism	1.57 (0.25)	6.45 (1.17)	4.27 (0.29)	1.57 (0.19)	−1.44 (0.20)
22. Compliance	1.29 (0.21)	4.52 (0.49)	2.19 (0.22)	−0.25 (0.14)	−2.78 (0.27)
23. Modesty	1.36 (0.22)	4.71 (0.61)	2.51 (0.26)	0.03 (0.15)	−2.37 (0.26)
24. Tender-mindedness	1.39 (0.25)	4.73 (0.58)	3.30 (0.35)	0.95 (0.17)	−1.52 (0.20)
Total information/IRT reliability		13.19 (62.27%)/.92
EIU (IU upper and lower bounds)		1.50 (1.49-2.94)
NMRU	.51
Conscientiousness
25. Competence	2.15 (0.27)	6.78 (1.11)	4.63 (0.44)	1.44 (0.22)	−2.31 (0.26)
26. Order	1.81 (0.21)	4.88 (0.53)	2.51 (0.23)	0.23 (0.16)	2.04 (0.22)
27. Dutifulness	2.42 (0.29)	7.61 (1.48)	4.65 (0.47)	1.46 (0.22)	−2.47 (0.29)
28. Achievement Striving	2.19 (0.27)	5.38 (0.68)	3.27 (0.37)	0.25 (0.18)	−3.22 (0.32)
29. Self-Discipline	2.41 (0.28)	7.18 (1.16)	4.00 (0.40)	0.78 (0.20)	−3.03 (0.31)
30. Deliberation	0.97 (0.15)	4.69 (0.63)	2.60 (0.24)	0.34 (0.14)	−1.73 (0.18)
Total information/IRT reliability		30.68 (81.21%)/.97
EIU (IU upper and lower bounds)		1.58 (1.56-2.55)
NMRU	.61

In our adolescent high school student sample, the full-information five-factor confirmatory IRT model showed a highly significant value for the M₂*statistic (i.e., a test statistic that can be used in the assessment of model fit in applications of IRT when the items are polytomous; Cai & Hansen, 2013), M₂*(315) = 1328.199, p < .001; however, RMSEA value suggested acceptable fit, RMSEA = .079 (95% CI: .075, .084). Similar results were observed in our adult community sample for both five-factor confirmatory IRT model, M₂*(315) = 1220.18, p < .001, RMSEA = .079 (95% CI: .075, .084). Finally, test information curves for the FFMRF domain scales in the adolescent sample and adult sample, respectively, are shown in Figure 1. Test information curves depicted the amount of information provided across levels of the latent trait; thus they can be used to reveal how informative a measure is at all levels of the latent trait (Baker, 2001). Across all five personality domains, the FFMRF scales covered a wide θ range; for instance, the Agreeableness dimension provided most information at a θ range of −3.0 to 1.0, in both adolescent sample and adult sample. Nonetheless, test information curves showed that the FFMRF domains provided more information when the FFMRF was administered to adult participants than adolescent participants (with the partial exception of Agreeableness domain).

Figure 1.

Test information curves for the FFMRF domain scales in the adolescent sample (continuous line), and in the adult sample (dotted line).

Categorical Regression Analysis and Multiple Regression Analysis Results

Item-level regression analyses (i.e., CATREG-LASSO models) in adolescent high school students and in community dwelling adults, respectively, are summarized in Table 4. For ease of presentation, only standardized regression coefficients for the FFMRF items that were included in the most parsimonious CATREG-LASSO model based on .632 bootstrap estimate are displayed. Standard error (SE) estimates were based on 1,000 bootstrap samples; β coefficient p values were based on CATREG-LASSO F tests.

Table 4.

Five Factor Model Rating Form Item-Level Regression Analyses in Adolescent Participants (N = 510) and in Adult Participants (N = 457): CATREG-LASSO Models Summary Table.

	Adolescent Sample (N = 510)		Adult Sample (N = 457)
	BPFSC-11	Boldness	Meanness	Disinhibition	TriPM Total
	β (SE)	β (SE)	β (SE)	β (SE)	b (SE)
1. Anxiousness	.03 (.03)	−.16*** (.04)	−.00 (.02)		−.05 (.03)
2. Angry Host.	.11*** (.04)	−01 (.03)	.21*** (.04)	.19*** (.04)	.14*** (.04)
3. Depressiveness	.22*** (.04)	−.02 (.03)
4. Self-consc.		−.08** (.04)			−.00 (.03)
5. Impulsivity	.09*** (.04)		.00 (.03)	.13*** (.04)	.11*** (.04)
6. Vulnerability	.15*** (.04)	−.17*** (.04)	−.02 (.03)		−.08**(.04)
7. Gregariousness			−.07 (.05)		−.02 (.04)
8. Assertiveness
9. Activity		.20*** (.04)			.07 (.04)
10. Excitement-S.
11. Positive Em.		.11*** (.04)	.13***(.04)	.08* (.04)	.21*** (.04)
12. Gregarious.	−.03 (.03)
13. Fantasy				.04 (.03)
14. Aesthetics		.04 (.03)	−.01 (.03)		−.01 (.03)
15. Feelings		.02 (.03)		−.03 (.04)
16. Actions		.10*** (.04)	.05 (.04)		.10*** (.04)
17. Ideas		.01 (.03)
18. Values
19. Trust					−.01 (.03)
20. Straightfor.			−.04 (.04)	−.01 (.03)	−.03 (.04)
21. Altruism			−.04 (.04)
22. Compliance			−.02 (.03)		−.02 (.03)
23. Modesty		−.08*** (.03)	−.02 (.03)		−.09** (.04)
24. Tender-mind.			−.17*** (.06)		−.04 (.04)
25. Competence		.06* (.03)		−.06 (.04)
26. Order	−.08** (.04)			−.03 (.04)
27. Dutifulness				−.04 (.04)	−.03 (.03)
28. Achievement		.06* (.03)
29. Self-Dis.				−.06 (.04)
30. Deliberation				−.03 (.04)
R ² _adjusted	.35***	.51***	.30***	.32***	.39***
R ² _regularized	.34***	.53***	.31***	.31***	.41***

Note. For ease of presentation, in each CATREG-LASSO regression equation only β coefficients for the FFMRF items that were included in the most parsimonious model based on .632 bootstrap estimate are displayed. BPFSC-11: Borderline Personality Features Scale-11; TriPM: Triarchic Personality Measure; SE estimates are based on 1,000 bootstrap samples; β coefficient p-values were based on CATREG-LASSO F-tests regularized R² was computed as 1-error.

p <.05; **p <.01; ***p <.001.

In BPFSC-11 analyses, variance inflation factor (VIF) values for the FFMRF items before LASSO regularization ranged from 1.23 to 1.70 (mean VIF value = 1.47, SD = 0.12), whereas after LASSO regularization they ranged from 1.15 to 1.51 (M = 1.33, SD = 0.10). According to the most parsimonious CATREG-LASSO model, seven FFMRF items explained more than 30% of the variance in the BPFSC-11 total score; in particular, FFMRF Angry Hostility (+), Depressiveness (+), Impulsivity (+), Vulnerability (+), and Orderlines (−) items significantly predicted the BPFSC-11 total score.

In the adult sample, when the TriPM Boldness scale was entered in the CATREG-LASSO model as dependent variable VIF values for the FFMRF items ranged from 1.27 to 2.09 (M = 1.61, SD = 0.22) before LASSO regularization, whereas after LASSO regularization they ranged from 1.13 to 1.69 (M = 1.41, SD = 0.17). Similar VIF values were observed when TriPM Meanness (VIF values before LASSO regularization: M = 1.61, SD = 0.23; VIF values after LASSO regularization: M = 1.44, SD = 0.25), Disinhibition (VIF values before LASSO regularization: M = 1.61, SD = 0.23; VIF values after LASSO regularization: M = 1.36, SD = 0.17), and total score (VIF values before LASSO regularization: M = 1.61, SD = 0.23; VIF values after LASSO regularization: M = 1.41, SD = 0.23), respectively were entered as dependent variables in CATREG-LASSO models.

The most parsimonious CATREG-LASSO models explained a nonnegligible amount of variance in the TriPM scores, ranging from roughly 30% (TriPM Meanness scale) to more than 50% (TriPM Boldness scales).

The results of multiple regression analyses¹ of the BPFSC-11 total score on the FFMRF domain scale scores in the adolescent sample, and of the TriPM scale scores and total score on the FFMRF domain scale scores in the adult sample are summarized in Table 5. To reduce the risk of capitalizing on chance, within each multiple regression equation the nominal significance level (i.e., p < .05) was corrected according to the Bonferroni procedure and set at p < .01. In both samples VIF values did not evidence any collinearity problem.

Table 5.

The Five Factor Model Rating Form Domain Scales as Predictors of the Borderline Personality Features Scale-11 in the Adolescent Sample (N = 510), and of the Triarchic Psychopathy Measure Scale Scores and Total Score in the Adult Sample (N = 457): Multiple Regression and Moderation Analysis Summary Table.

Five Factor Model Rating Form	Adolescent Sample		Adult Sample
	BPFSC-11 Total Score		TriPM Boldness		TriPM Meanness		TriPM Disinhibition		TriPM Total Score
	β	VIF	β	VIF	β	VIF	β	VIF	β	VIF
Neuroticism	.51	1.14	−.32	1.13	.12	1.13	.33	1.13	.05	1.13
Extraversion	−.05	1.33	.31	1.40	.06	1.40	.16	1.39	.26	1.39
Openness	.14	1.21	.25	1.23	.08	1.23	.10	1.23	.20	1.23
Agreeableness	−.05^a	1.15	−.29	1.10	−.40^b	1.10	−.11	1.10	−.38	1.10
Conscientiousness	−.16	1.11	.10	1.14	−.05	1.14	−.32	1.14	−.12	1.14
R ² _adjusted	.34***		.41***		.17***		.24***		.22***

Note. BPFSC-11: Borderline Personality Features Scale-11; TriPM: Triarchic Psychopathy Measure; VIF: Variance inflation factor. Within each regression equation the nominal significance level (i.e., p <.05) for the regression coefficients was corrected according to the Bonferroni procedure and set at p <.01. Bold highlights standardized regression coefficient that were significant at Bonferroni-corrected p-level. ***p <.001.a: gender-by-A interaction: R² = .01, β = -.07, p <.05, male adolescents’ β = .03, p >.10, female adolescents’ β = -.12, p <.05; b: gender-by-A interaction: R² = .01, β = .11, p <.05, male adults’ β = .-.44, p <.001, female adults’ β = -.23, p <.001.

In the adolescent sample, the FFMRF scales explained a non-negligible amount of variance in the BPFSC-11 total score; N (+), O (+), and C (−) were the most significant predictors of the BPFSC-11 total score. In the adult sample, the FFMRF domain scales predicted a nontrivial amount of variance in all TriPM dimensions, particularly in TriPM Boldness scale. N (−), E (+), O (+), and A (−) significantly predicted TriPM Boldness scores, whereas low A was the strongest predictor of TriPM Meanness scores with a marginal, albeit significant contribution of high N; TriPM Disinhibition was significantly predicted by C (−), N (+), and E (+). Finally, the TriPM total score was significantly predicted by A (−), E (+), O (+), and C (−).

Discussion

As a whole, our findings do appear to have confirmed as well as to extend to a different culture (i.e., Italian society) and age range (i.e., adolescence) the available evidence on the reliability and validity of the FFMRF, at least in its self-report version. On average, moderate Cronbach α values were observed both in the adolescent sample (median Cronbach α value = .66, SD = .06) and in the adult sample (median Cronbach α value = .67, SD = .07); these values were highly consistent with previous reports on the internal consistency of the FFMRF scales in U.S. samples (Mullins-Sweatt et al., 2006; Mullins-Sweatt & Widiger, 2010; Samuel et al., 2013). It should be observed that in our study average interitem r values were in the .15 to .50 range for all FFMRF scales in both samples, suggesting that the FFMRF domain scales may have adequate reliability (Clark & Watson, 1995). Indeed, Cronbach α coefficient is consistent with domain sampling model assumptions, it may be undesirable in the case of scales including a very limited number of items (Clark & Watson, 1995), like the FFMRF domain scales.

IRT reliability estimates seemed to support these concerns on using reliability indices based on domain sampling model assumptions in the case of measures that were developed according to different models, including a very limited number of items in each scale. Indeed, in our study IRT reliability index values were equal to, or greater than .90, for all FFMRF domain scales in both adolescent participants and adult participants. These findings suggested that the FFMRF scales yielded adequately reliable scores for all FFM domains in both community dwelling adolescents and adults. In other words, Markon’s (2013a) information utility index values were highly consistent with conventional IRT reliability estimates in suggesting that the FFMRF domain scale scores were provided with adequate reliability, at least in the current samples. Although further research is always needed to better establish what is typical for information utility indices and how to interpret them on an absolute scale (Markon, 2013a), for all FFMRF domains in the current study information utility lower bounds were greater than 1.2, information utility upper bounds were greater than 1.3, and NMRU values were greater than .45. These findings are consistent with those reported in Markon’s (2013a) study for the full length Schedule for Nonadaptive and Adaptive Personality (Clark, Simms, Wu, & Casillas, in press) and Dimensional Assessment of Personality Pathology (Livesley & Jackson, 2009) scales which represent extremely well-validated, reliable measures that have been used in numerous contexts. Although it is somewhat unclear how well these might generalize to other settings, Markon (2013a) suggested that these values might provide heuristic guidelines for tests with similar characteristics.

Full-information IRT analysis results supported the five-dimensional model of FFMRF items. Indeed, our full-information confirmatory IRT analysis based on the a priori FFM of the FFMRF items showed acceptable fit in both adolescent sample and adult sample. In terms of item validity, the wide majority of FFMRF facet scores showed substantial discrimination parameters, showing that they were related to the corresponding FFM domain in a way that was consistent with the expectations of NEO Five Factor Theory (Costa & McCrae, 1995). However, a noticeable variability in validity parameter estimates and item “difficulty” values was observed. In our adolescent sample, 80.0% (n = 24) of the FFMRF items showed substantial and significant (i.e., ≥ 0.90) slope parameter values—that is, parameter values which represent the degree of relation of item responses to the underlying construct—whereas only one item (Item 5, “Impulsivity; tempted, urgency vs. controlled, restrained”) showed a weak (i.e., slope parameter value <.50) albeit significant relationship with the underlying dimension. In the community dwelling adult sample, 26 FFMRF items (86.7%) showed discrimination (i.e., slope parameter) values that were significant and equal to, or greater than 0.90, whereas no FFMRF item showed weak relationship with the corresponding latent dimension—Item 5 from Neuroticism showed the lowest discrimination value among FFMRF items in the adult sample. Although this finding may reflect relatively low specificity of FFMRF Item 5 with respect to the N domain, local dependence analyses supported the unidimensionality hypothesis for all FFMRF items. According to the IRT findings, FFMRF Item 5 did not seem to adequately capture the latent distribution of the Impulsivity facet of N. Rather than casting doubts on the relationships between N and Impulsivity or hypothesizing multidimensionality of Impulsivity facet, we feel that our IRT analysis findings raised issues as to the adequacy of adjective-based measures in adequately capturing the essence of the Impulsivity facet of N while discriminating it from other sources of behavior dyscontrol. Indeed, “tempted, urgency” versus “controlled, restrained” (i.e., FFMRF Item 5 adjectives) may not efficiently help participants—particularly adolescent participants—to discriminate so-called Negative Urgency (i.e., behavior dyscontrol due to activation of negative emotionality) from other sources of behavior dyscontrol (e.g., lack of perseverance, lack of planning, and sensation seeking), which are not related to high N in terms of FFM domains (Cyders & Smith, 2007).

Test information curves showed that FFMRF domains provided more information when the FFMRF was administered to adult participants than adolescent participants; this finding is consistent with previous studies (e.g., Spence, Owens, & Goodyer, 2012) showing that personality traits are still in flux throughout adolescence (McCrae et al., 2002). Although a higher proportion of FFMRF items with discrimination parameter values greater than .90 was observed among community dwelling adults than among community dwelling adolescents, the difference did not reach statistical significance (i.e., p > .60) suggesting that the FFMRF items performed roughly the same in the adolescent sample and in the adult sample. In summary, the IRT analysis findings yield at least moderate support to the hypothesis that the FFMRF represent a reliable and valid measure of the corresponding FFM domains and facets, which can be safely used both in community dwelling adolescents and adults.

As a whole, regression analysis results seemed to support the possible clinical usefulness of the FFMRF scores, at least with respect to borderline personality features and psychopathy characteristics in community dwelling adolescents and adults, respectively. Many of the associations that were observed in our study between FFMRF scores and measures of borderline personality features and psychopathy traits were somewhat consistent with the nomological network for measures of FFM facets and traits (Poy et al., 2014; Samuel & Widiger, 2008).

In item-level analyses, CATREG-LASSO models showed the primacy of the FFMRF N items in predicting the level of borderline personality pathology in nonclinical adolescents, at least as it is operationalized in the BPFSC-11 total score. This finding was in close agreement with Samuel and Widiger’s (2008) meta-analytic results, although we were not able to replicate the negative associations between borderline personality disorder, and selected A and E facets. We feel that three major factors may help explaining this partial difference in the FFM facet profile: (a) Samuel and Widiger’s (2008) meta-analysis included studies using measures based on DSM-IV (American Psychiatric Association, 2013) criteria for assessing borderline personality disorder; rather, we relied on a scale which was designed to yield a dimensional assessment of borderline personality pathology without capitalizing on a psychiatric approach; (b) all studies listed in Samuel and Widiger’s (2008) meta-analysis included the NEO-PI-R among the FFM measures, whereas we relied solely on the FFMRF, which provides a single item for each facet; (c) it is possible that age differences may play a role in shaping the associations between borderline personality pathology and selected FFM domain facets.

In our study, CATREG-LASSO analysis results indicated that FFMRF facets explained a non-negligible amount of variance in all TriPM dimensions as well as in the TriPM total score. In particular, CATREG-LASSO results suggested that low scores on Anxiousness (N), Self-consciousness (N), Vulnerability (N), and Modesty (A), and high scores on Activity (E), Positive Emotions (E), Actions (O), Competence (C), and Achievement Striving (C) were significant predictors of Boldness; this finding was at least partially consistent with Poy et al.’s (2014) results, although we relied on a multivariate approach in our facet-level analyses rather than on a bivariate correlation approach as in Poy et al.’s (2014) study.

Considering Meanness, CATREG-LASSO analysis results showed that it was significantly predicted by low scores on Tender-mindedness (A), and high scores on Angry Hostility (N) and Positive Emotions (E). This results suggested that subjects scoring high on Meanness “feel good about being bad,” that is, they seem to have positive feelings about being callous and hostile; although this result is largely consistent with Patrick et al.’s (2009) construct of Meanness, it should be observed that it did not completely overlap with Poy et al.’s (2014) findings, particularly with respect to the direction of the association between Positive Emotions and Meanness.

Similar considerations held also for our findings concerning the relationship between FFMRF facet scores and TriPM Disinhibition scale scores. In our study, high scores on Angry Hostility (N), Impulsivity (N), and Positive Emotions (E) significantly predicted Disinhibition. Interestingly, all facets of C were included in our CATREG-LASSO model for Disinhibition and showed a negative sign, although none of them reached statistical significance; this finding seemed to suggest that low scores on C may be relevant in defining the Disinhibition dimension of psychopathy, with no specific contribution of the individual facets (total FFMRF C did relate well to Disinhibition). This consideration is also consistent with Poy et al.’ (2014) findings, which showed a negative association between Disinhibition and Competence, Dutifulness, Self-Discipline, and Deliberation (i.e., facets of C), respectively.

Multiple regression analyses based on FFMRF domain scores yielded further support for the potential clinical usefulness of the FFMRF in community dwelling participants. In our adolescent sample, the FFMRF domain scale predicted a substantial amount of variance in the BPFSC-11 total score. Interestingly, the association between the BPFSC-11 total score and high N and low C, respectively, was largely consistent with the FFM domain profile for borderline personality disorder which has been documented in Samuel and Widiger’s (2008) meta-analysis.

Consistent with the net of relationships that was identified by Poy et al. (2014) using the NEO-PI-R as measure of FFM domains, in our adult sample Boldness was significantly predicted by low N, high E, high O, and low A. Moreover, in our study Disinhibition was significantly predicted by C (−), N (+), and E (+); with the exception of the association between low A and Disinhibition, our findings were in close agreement with Poy et al.’s (2014) results. Extending available evidence on the relationship between FFM domains and the triarchic construct of psychopathy, in our study A (−), E (+), O (+), and C (−) were significant predictors of the overall level of psychopathy as it is operationalized in the TriPM total score. Although further studies are needed before accepting our findings, our results suggest that community dwelling adults scoring high on the overall level of triarchic construct of psychopathy appear as fearless (i.e., low Vulnerability), boastful (i.e., low Modesty), high-spirited (i.e., high Positive Emotions), unrestrained and with high urgency (i.e., high Impulsivity), and angry (i.e., high Angry Hostility); it should be observed that this finding were consistent with Patrick et al.’s (2009) description of the personality features characterizing psychopathy according to the triarchic model.

In summary, we feel that our regression analysis findings suggested that the FFMRF may have potential clinical usefulness, at least in screening for borderline personality pathology and psychopathy in community dwelling adolescents and in community dwelling adults, respectively. Indeed, the FFMRF is a short measure that can be administered and scored in a few minutes and is available with no economic charge; moreover, the FFMRF items do not contain stigmatizing language or psychiatric terms. Thus, the FFMRF may represent a promising instrument for clinicians who need to identify adolescents at risk for borderline personality pathology in community settings (e.g., psychologists working in schools, pediatric care professionals, etc.), or to screen for psychopathy risk among community dwelling adults (e.g., general practitioners, human resources psychologists, etc.). It should be observed that the FFMRF, as with all FFM measures, yields personality profiles, which do not simply indicate personality dysfunction, but also personality resilience factors; this may represent another appealing feature for clinicians who are interested in a short measure of screening based on the FFM.

Of course, the results of our study should be considered in the light of several limitations. One of the major limitations of the present study involves the use of different self-report measures (i.e., the BPFSC-11 in the adolescent sample and the TriPM in the adult sample) to evaluate the potential usefulness of the FFMRF in a nomological network perspective. Moreover, it should be stressed that the BPFSC-11 is not based on DSM-5 Section II (or even Section III) criteria for borderline personality disorder; using DSM-based measures to assess borderline personality disorder in adolescents may yield different results. Similar considerations held also for the TriPM; there is no evidence that using different measures of psychopathy as criterion variable will not affect the relationships with the FFMRF domain scores and facet scores. For example, whereas TriPM Boldness relates strong to FFM neuroticism and extraversion, this is not a finding that is obtained with more traditional measures of psychopathy (Crego & Widiger, 2014).

In the present study, we relied on a very limited set of pathological personality traits (namely, borderline personality disorder and psychopathy) that were measured using the same method (i.e., self-report questionnaires); further evidence of the predictive validity of the FFMRF scores with respect to a wide range of non-adaptive personality traits, possibly assessed using methods different from self-reports, is needed before drawing definite conclusion on the FFMRF as a clinically useful measure of the FFM traits. These considerations further emphasize that replication and extension to other external measures is mandatory before accepting our findings.

In our regression analyses we relied on self-report measures to assess both predictive variables (i.e., the FFMRF scales and items) and criterion variables (i.e., BPFSC-11 total score and TriPM scale/total scores, respectively). Thus, shared method variance may have spuriously increased the nomological network validity coefficients of the FFMRF domain scales and facet scores.

Although we demonstrated that the FFMRF scores significantly and meaningfully predict clinically relevant constructs, we relied solely on adolescent and adult participants dwelling in the Italian community; it is unclear how well our findings can be extended to clinical populations. In addition, none of our samples was composed of randomly selected participants; notwithstanding their moderately large size, they were more akin to convenient study groups than to samples of participants actually representative of the Italian population.

As a whole, these considerations limit the generalizability of our findings and stress the needs for further studies before finally accepting our conclusions. Even keeping these limitations in mind, we feel that our findings provide further evidence of the reliability and validity of the FFMRF, extending available evidence to different developmental phases (adolescence and adulthood) and cultures, and highlighting the clinical potential of the FFMRF in screening for selected personality pathology (namely, borderline personality disorder in adolescents and psychopathy in adults).

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.

Axelrod

S. R.

Widiger

T. A.

Trull

T. J.

Corbitt

E. M.

(1997). Relation of five-factor model antagonism facets with personality disorder symptomatology. Journal of Personality Assessment, 69, 297-313.

Baker

(2001). The basics of item response theory. College Park, MD: ERIC Clearinghouse on Assessment and Evaluation, University of Maryland.

Briggs

S. R.

(1992). Assessing the five-factor model of personality description. Journal of Personality, 60, 253-293.

Cai

Hansen

(2013). Limited-information goodness-of-fit testing of hierarchical item factor models. British Journal of Mathematical and Statistical Psychology, 66, 245-276.

Cai

Maydeu-Olivares

Coffman

D. L.

Thissen

(2006). Limited information goodness-of-fit testing of item response theory models for sparse 2^p tables. British Journal of Mathematical and Statistical Psychology, 59, 173-194.

Caldwell

M. F.

Ziemke

M. H.

Vitacco

M. J.

(2008). An examination of the Sex Offender Registration and Notification Act as applied to juveniles: Evaluating the ability to predict sexual recidivism. Psychology, Public Policy, and Law, 14, 89-114.

Cha

E. S.

Kim

K. H.

Erlen

J. A.

(2007). Translation of scales in cross-cultural research: Issues and techniques. Journal of Advance Nursing, 58, 386-395.

Chalmers

R. P.

(2012). mirt: A multidimensional item response theory package for the R environment. Journal of Statistical Software, 48, 1-29.

10.

Chen

W. H.

Thissen

(1997). Local dependence indices for item pairs using item response theory. Journal of Educational and Behavioral Statistics, 22, 265-289.

11.

Clark

L. A.

Simms

L. J.

K. D.

Casillas

(in press). Schedule for Nonadaptive and Adaptive Personality—Second Edition (SNAP-2). Minneapolis: University of Minnesota Press.

12.

Clark

L. A.

Watson

(1995). Constructing validity: Basic issues in objective scale development. Psychological Assessment, 7, 309-319.

13.

Clarke

Barron

A. R.

(1994). Jeffreys’ prior is asymptotically least favorable under entropy risk. Journal of Statistical Planning and Inference, 41, 37-60.

14.

Costa

P. T.

Jr. McCrae

R. R.

(1992). Revised NEO Personality Inventory (NEO PI-R) and NEO Five-Factor Inventory (NEO-FFI) professional manual. Odessa, FL: Psychological Assessment Resources.

15.

Costa

P. T.

Jr. McCrae

R. R.

(1995). Domains and facets: Hierarchical personality assessment using the Revised NEO Personality Inventory. Journal of Personality Assessment, 64, 21-50.

16.

Crego

Widiger

T. A.

(2014). Psychopathy, DSM-5, and a caution. Personality Disorders: Theory, Research, and Treatment, 5, 335-347.

17.

Crick

N. R.

Murray-Close

Woods

(2005). Borderline personality features in childhood: A short-term longitudinal study. Development and Psychopathology, 17, 1051-1070.

18.

Cyders

M. A.

Smith

G. T.

(2007). Mood-based rash action and its components: Positive and negative urgency. Personality and Individual Differences, 43, 839-850.

19.

Denissen

J. J.

Geenen

van Aken

M. A.

Gosling

S. D.

Potter

(2008). Development and validation of a Dutch translation of the Big Five Inventory (BFI). Journal of Personality Assessment, 90, 152-157.

20.

DeYoung

C. G.

Quilty

L. C.

Peterson

J. B.

(2007). Between facets and domains: 10 aspects of the Big Five. Journal of Personality and Social Psychology, 93, 880-896.

21.

Drislane

L. E.

Patrick

C. J.

Arsal

(2014). Clarifying the content coverage of differing psychopathy inventories through reference to the triarchic psychopathy measure. Psychological Assessment, 26, 350-362.

22.

Edelen

M. O.

Reeve

B. B.

(2007). Applying item response theory (IRT) modeling to questionnaire development, evaluation, and refinement. Quality of Life Research, 16, 5-18.

23.

Efron

(1983). Estimating the error rate of a prediction rule: Improvement on cross-validation. Journal of the American Statistical Association, 78, 316-331.

24.

Embretson

S. E.

Reise

S. P.

(2000). Item response theory for psychologists. Mahwah, NJ: Erlbaum.

25.

Forero

C. G.

Vilagut

Adroher

N. D.

Alonso

ESEMeD/MHEDEA Investigators. (2013). Multidimensional item response theory models yielded good fit and reliable scores for the Short Form-12 questionnaire. Journal of Clinical Epidemiology, 66, 790-801.

26.

Fossati

Sharp

Borroni

Somma

(2015). Psychometric properties of the Borderline Personality Features Scale for Children-11 (BPFSC-11) in a sample of community dwelling Italian adolescents. Manuscript submitted for publication.

27.

Geisinger

K. F.

(1994). Cross-cultural normative assessment: Translation and adaptation issues influencing the normative interpretation of assessment instruments. Psychological Assessment, 6, 304-312.

28.

Ghosh

Mergel

Liu

(2011). A general divergence criterion for prior selection. Annals of the Institute of Statistical Mathematics, 63, 43-58.

29.

Gretton

H. M.

Hare

R. D.

Catchpole

R. E.

(2004). Psychopathy and offending from adolescence to adulthood: A ten year follow-up. Journal of Clinical and Consulting Psychology, 72, 636-645.

30.

Hattie

(1985). Methodology review: Assessing dimensionality of tests and items. Applied Psychological Measurement, 9, 139-164.

31.

Hoerl

A. E.

Kennard

(1970a). Ridge regression: Applications to nonorthogonal problems. Technometrics, 12, 69-82.

32.

Hoerl

A. E.

Kennard

(1970b). Ridge Regression: biased estimation for nonorthogonal problems. Technometrics, 12, 55-67.

33.

Howell

A. J.

Dopko

R. L.

Turowski

J. B.

Buro

(2011). The disposition to apologize. Personality and Individual Differences, 51, 509-514.

34.

Jeffreys

(1946). An invariant form for the prior probability in estimation problems. Proceedings of the Royal Society A: Mathematical, Physical, & Engineering Sciences, 186, 453-461.

35.

Kennealy

P. J.

Hicks

B. M.

Patrick

C. J.

(2007). Validity of the factors of the Psychopathy Checklist–Revised in female prisoners. Assessment, 14, 323-340.

36.

Lane

Stone

(2006). Performance assessment. In Brennan

R. L.

(Ed.), Educational measurement (4th ed., pp. 387-431). Westport, CT: American Council on Education and Praeger.

37.

Lee

Ashton

M. C.

(2004). Psychometric properties of the HEXACO Personality Inventory. Multivariate Behavioral Research, 39, 329-358.

38.

Livesley

W. J.

Jackson

(2009). Manual for the Dimensional Assessment of Personality Pathology–Basic Questionnaire. Port Huron, MI: Sigma Press.

39.

Lynam

D. R.

Widiger

T. A.

(2001). Using the five-factor model to represent the DSM-IV personality disorders: An expert consensus approach. Journal of Abnormal Psychology, 110, 401-412.

40.

Markon

K. E.

(2013a). Information utility: Quantifying the total psychometric information provided by a measure. Psychological Methods, 18, 15-35.

41.

Markon

K. E.

(2013b). infutil: Information utility. R package version 1.0.

42.

McCrae

R. R.

Costa

P. T.

(2008). The five factor theory of personality. In John

O. P.

Robins

R. W.

Pervin

L. A.

(Eds.), Handbook of personality (3rd ed., pp. 159-181). New York, NY: Guilford.

43.

McCrae

R. R.

Costa

P. T.

Jr. Terracciano

Parker

W. D.

Mills

C. J.

De Fruyt

Mervielde

(2002). Personality trait development from age 12 to age 18: Longitudinal, cross-sectional and cross-cultural analyses. Journal of Personality and Social Psychology, 83, 1456-1468.

44.

Mullins-Sweatt

S. N.

Jamerson

J. E.

Samuel

D. B.

Olson

D. R.

Widiger

T. A.

(2006). Psychometric properties of an abbreviated instrument of the five-factor model. Assessment, 13, 119-137.

45.

Mullins-Sweatt

S. N.

Widiger

T. A.

(2010). Personality-related problems in living: An empirical approach. Personality Disorders: Theory, Research, and Treatment, 1, 230-238.

46.

Nunnally

J. C.

Bernstein

I. H.

(1994). Psychometric theory (3rd ed.). New York, NY: McGraw-Hill.

47.

Patrick

C. J.

(2010). Operationalizing the triarchic conceptualization of psychopathy: Preliminary description of brief scales for assessment of boldness, meanness, and disinhibition (Unpublished test manual). Tallahassee: Florida State University. Test retrieved from https://www.phenxtoolkit.org/index.php?pageLink=browse.protocoldetails&id=121601

48.

Patrick

C. J.

Drislane

L. E.

(2015). Triarchic model of psychopathy: Origins, operationalizations, and observed linkages with personality and general psychopathology. Journal of Personality, 83, 627-643. doi:10.1111/jopy.12119

49.

Patrick

C. J.

Fowles

D. C.

Krueger

R. F.

(2009). Triarchic conceptualization of psychopathy: Developmental origins of disinhibition, boldness, and meanness. Development and Psychopathology, 21, 913-938.

50.

Paunonen

S. V.

Ashton

M. C.

(2001). Big five factors and facets and the prediction of behavior. Journal of Personality and Social Psychology, 81, 524-539.

51.

Paunonen

S. V.

Haddock

Fosterling

Keinonen

(2003). Broad versus narrow personality measures and the prediction of behaviour across cultures. European Journal of Personality, 17, 413-433.

52.

Poy

Segarra

Esteller

À.

López

Moltó

(2014). FFM description of the triarchic conceptualization of psychopathy in men and women. Psychological Assessment, 26, 69-76.

53.

R Core Team. (2015). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved from http://www.R-project.org/

54.

Reise

Waller

(2002). Item response theory for dichotomous assessment data. In Drasgow

Schmitt

(Eds.), Measuring and analyzing behavior in organizations: Advances in measurement and data analysis (pp. 88-122). San Francisco, CA: Jossey-Bass.

55.

Reynolds

S. K.

Clark

L. A.

(2001). Predicting dimensions of personality disorder from domains and facets of the Five-Factor Model. Journal of Personality, 69, 199-222.

56.

Samejima

(1997). Graded response model. In van der Linden

W. J.

Hambleton

R. K.

(Eds.), Handbook of item response theory (pp. 85-100). New York, NY: Springer-Verlag.

57.

Samuel

D. B.

Mullins-Sweatt

S. N.

Widiger

T. A.

(2013). An investigation of the factor structure and convergent and discriminant validity of the five-factor model rating form. Assessment, 20, 24-35.

58.

Samuel

D. B.

Simms

L. J.

Clark

L. A.

Livesley

W. J.

Widiger

T. A.

(2010). An item response theory integration of normal and abnormal personality scales. Personality Disorders: Theory, Research, and Treatment, 1, 5-21.

59.

Samuel

D. B.

Widiger

T. A.

(2008). A meta-analytic review of the relationships between the five-factor model and DSM-IV-TR personality disorders: A facet level analysis. Clinical Psychology Review, 28, 1326-1342.

60.

Samuel

D. B.

Widiger

T. A.

(2011). Clinicians’ use of personality disorder models within a particular treatment setting: A longitudinal comparison of temporal consistency and clinical utility. Personality and Mental Health, 5, 12-28.

61.

Saucier

Goldberg

L. R.

(2002). Assessing the Big Five: Applications of 10 psychometric criteria to the development of marker scales. In De Raad

Perugini

(Eds.), Big five assessment (pp. 29-58). Bern, Switzerland: Hogrefe & Huber.

62.

Schenk

A. M.

Ragatz

L. L.

Fremouw

W. J.

(2012). Vicious dogs part 2: Criminal thinking, callousness, and personality styles of their owners. Journal of Forensic Sciences, 57, 152-159.

63.

Sellbom

Phillips

T. R.

(2013). An examination of the triarchic conceptualization of psychopathy in incarcerated and nonincarcerated samples. Journal of Abnormal Psychology, 122, 208-214.

64.

Sharp

Bleiberg

(2007). Borderline personality disorder in children and adolescents. In Martin

Volkmar

(Eds.), Lewis’ child and adolescent psychiatry: Comprehensive textbook (pp. 680-691). Baltimore, MD: Lippincott Williams & Wilkins.

65.

Sharp

Steinberg

Temple

Newlin

(2014). An 11-item measure to assess borderline traits in adolescents: Refinement of the BPFSC using IRT. Personality Disorders: Theory, Research, and Treatment, 5, 70-78.

66.

Sica

Drislane

Caudek

Angrilli

Bottesi

Cerea

Ghisi

(2015). A test of the construct validity of the Triarchic Psychopathy Measure in an Italian community sample. Personality and Individual Differences, 82, 163-168.

67.

Spence

Owens

Goodyer

(2012). Item response theory and validity of the NEO-FFI in adolescents. Personality and Individual Differences, 53, 801-807.

68.

Thissen

(1990). Reliability and measurement precision. In Wainer

(Ed.), Computerized adaptive testing: A primer (pp. 161-186). Hillsdale, NJ: Erlbaum.

69.

Thomas

K. M.

Yalch

M. M.

Krueger

R. F.

Wright

A. G. C.

Markon

K. E.

Hopwood

C. J.

(2013). The convergent structure of DSM-5 personality trait facets and Five-Factor Model trait domains. Assessment, 20, 308-311.

70.

Tibshirani

R. J.

(1996). Regression shrinkage and selection via the Lasso. Journal of the Royal Statistical Society Series B, 58, 267-288.

71.

Van

Vijver

F. J. R.

Hambleton

R. K.

(1996). Translating tests: Some practical guidelines. European Psychologist, 1, 89-99.

72.

van der Kooij

Meulman

J. J.

(2004). Regression with optimal scaling. In Meulman

J. J.

Heiser

W. J.

(Eds.), SPSS categories 13.0 (pp. 107-157). Chicago, IL: SPSS.

73.

van der Kooij

A. J.

Meulman

J. J.

Heiser

W. J

. (2006). Local minima in categorical multiple regression. Computational Statistics & Data Analysis, 50, 446-462.

74.

Walters

G. D.

Knight

R. A.

Grann

Dahle

(2008). Incremental validity of the Psychopathy Checklist facet scores: Predicting release outcome in six samples. Journal of Abnormal Psychology, 117, 396-405.

75.

Yen

(1984). Effects of local item dependence on the fit and equating performance of the three parameter logistic model. Applied Psychological Measurement, 8, 125-145.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.40 MB