Using the Self-Directed Search in Research

Abstract

Using Item Response Theory (IRT) and Confirmatory Factor Analysis (CFA), the goal of this study was to select a reduced pool of items from the French Canadian version of the Self-Directed Search—Activities Section (Holland, Fritzsche, & Powell, 1994). Two studies were conducted. Results of Study 1, involving 727 French Canadian students, showed that the psychometric qualities of the 66-item French Canadian version are equivalent to those of the original English version. Based on IRT and factor loadings derived from a CFA, 24 items were selected from the original 66 items. In Study 2 (n = 339 French Canadian young adults), we tested and obtained support for the construct validity of the 24 selected items using CFA and correlational analyses among interests’ dimensions. We concluded that the selected pool of items accurately captured Holland’s theoretical framework and showed adequate psychometrics qualities and construct validity.

Keywords

items selection self-directed search vocational interests Holland’s theory

Making a career choice is an important step in human development (Gati & Asher, 2001; Vondracek & Porfeli, 2003), given the significant consequences that it can have on personal and professional lives (Guay, Ratelle, Senécal, Larose, & Deschênes, 2006). Vocational interests, which have been found to be important determinants of career choice (Lent, Brown, & Hackett, 2002; Osipow & Fitzgerald, 1996), are defined as patterns of personal likes, dislikes, and indifferences regarding various career-relevant activities (Lent et al., 2002). Multiple inventories have been developed to assess vocational interests (Guglielmi, Fraccaroli, & Pombeni, 2004; Hood & Johnson, 2007; Kapes, Mastie, & Whitfield, 1994), most of which are based on Holland’s theory of vocational choices (Holland, 1992, 1997).

Holland’s theory (1992, 1997) states that most people can be classified according to six personality types: Realistic, Investigative, Artistic, Social, Enterprising, and Conventional (RIASEC). Each of these types is characterized by specific beliefs, values, abilities, and interests. These six personality types are theoretically situated on a hexagon that represents the expected relations among them (Holland, 1997). For example, two adjacent types on the hexagon, such as Realistic and Investigative, share more characteristics than two opposite types, such as Artistic and Conventional.

The Self-Directed Search ([SDS] Holland, Fritzsche, & Powell, 1994), based on Holland’s theory, is one of the most widely used interest inventories because it can be self-administered and self-interpreted, and the scoring does not require electronic tools (Spokane & Holland, 1995). In addition, the factorial structure of the SDS obtained empirical support (Khan & Alvi, 1991; Oosterveld, 1994). Finally, the SDS has been translated in 20 different languages (see Glidden-Tracey & Greenwood, 1997; Muck, 2005; Yang, Lance, & Hui, 2006 for examples), with one of the versions being a French Canadian one (Holland, 1991).

Like its original version, the French Canadian version of the SDS contains 228 items divided into 4 sections (Activities, Competencies, Occupations, and Self-Estimates) measuring several career choice determinants. Vocational interests are measured by the Activities section of this inventory, in which participants indicate whether they would like to do each of the 66 proposed career activities. These activities can all be categorized according to the six personality dimensions. Summing the scores obtained on each dimension yields a measure of the participant’s vocational interest profile.

Counselors usually have sufficient time to administer the SDS-Activities section to the counselee. Unfortunately, when it comes to research, investigators are usually limited by time such that they cannot administer lengthy inventories like the SDS, in addition to assessing other important constructs. Because researchers generally aim to further their understanding of the determinants or consequences of career interests, they need to include other measures in their survey. Knowing that longer surveys are expected to affect the response rate and amount of missing data (Graham, 2009), researchers often seek out a reduced pool of items with adequate psychometric properties. Hence, with 66-item, the SDS-Activities scale can be relatively long for research purposes. Despite the fact that using a restricted pool of items from a longer scale typically tends to yield worst psychometric qualities than longer versions, they have the advantage of diminishing item redundancy, which in turn reduces participants’ fatigue, frustration, and boredom (Robins, Hendin, & Trzesniewski, 2001). Furthermore, they can be considered to have acceptable, albeit not as high, psychometric qualities (e.g., factor loadings above .50, Cronbach’s αs .70 and higher, etc.). In fact, selecting a pool of items from an original scale has gained in popularity over the past few years. Notably, this is the case for the Inventory of Children’s Individual Differences (Deal, Halverson, Martin, Victor, & Baker, 2007), the Big Five Factor (Gosling, Rentfrow, & Swann, 2003; Muck, Hell, & Gosling, 2007; Rammstedt & John, 2007) and the Dyadic Adjustment Scale (Sabourin, Valois, & Lussier, 2005).

The Present Research

The main goal of the present research was to select a reduced pool of items from the French Canadian version of the Self-Directed Search—Activities section (Holland, 1991). The selected pool of items has to be small enough to allow a rapid assessment of vocational interest but also have sufficient items to ensure more accurate parameter estimates and greater reliability (Marsh, Hau, Balla, & Grayson, 1998). Based on Marsh et al.’s (1998) recommendation that a factor should include a minimum of 4 items, our goal was to select a total of 24 items (i.e., 4 items for each RIASEC dimension). This goal was pursued in two studies. Study 1 comprised two steps: (a) we verified whether the psychometric properties of the 66-item French Canadian version were equivalent to those of the English version, and (b) we selected a pool of items from the 66-item French Canadian version using Item Response Theory (IRT) and factor loadings from a Confirmatory Factor Analysis (CFA). Study 2 comprised three steps: (a) we tested the construct validity of the selected pool of items via a CFA, (b) we examined whether the factorial structure of the selected pool of items was invariant across gender (see Guglielmi et al., 2004; Turner et al., 2008), and (c) we verified if the correlations among subscales of this pool were comparable to those of the 66-item English version, as reported in the SDS technical manual (Holland et al., 1994).

The advantages of this research are twofold. First, it will provide useful information on the cross-cultural validity of the French Canadian version of the SDS, which, to the best of our knowledge, has never been assessed. Second, the validity and reliability of the selected items could be further tested by researchers working with the 66-item English-speaking populations as well as those speaking other languages (e.g., Spanish, Arabic, Vietnamese, and Chinese; Glidden-Tracey & Greenwood, 1997; Muck, 2005; Yang et al., 2006 for examples), especially when researchers do not have enough time to administer the full-length version.

Study 1

Method

Participants

A total of 727 French Canadian students (211 men, 511 women, and 5 unspecified) participated in the study. Participants’ mean age was 21.43 years (SD = 4.11; range = 16−51 with 38 participants being over 30). Of the 727 participants, 490 were university students and 237 attended college. In the province of Quebec, high school graduates can attend college in either a 2-year pre-university program or a 3-year technical program, leading to the labour market. A total of 0.08% of college students had academic achievement scores above 90%, 91.9% had scores between 70% and 80%, and 8.02% had scores lower than 70%. Among university students, 8.3% had academic achievement scores above 90%, 87.1% had scores between 70% and 80%, and 4.6% had scores lower than 70%. Overall, 94.3% of the participants were born in the province of Quebec, 1.4% in another Canadian province, and 4.3% in another country. A total of 97.2% of them spoke French as their first language, 1.1% had English as a first language, and 1.7% spoke another language. Half of the college students were in their third semesters (50.6%), while the remaining (49.4%) were in their first, second, fourth, fifth, or sixth semesters. Among university students, 27.8% were in their first semester, 29% in their third semester, whereas 43.2% were in their second, fourth, fifth, sixth, seventh, eighth, or ninth semester. In terms of domain, 45% of college students studied social sciences or humanities, 20.5% arts, and 34.5% other areas. Among university students, 35.6% majored in psychology, 21.8% in school guidance and counselling, and 42.6% in other programs. A total of 34.5% of the participants lived with both parents, 23.1% lived with one parent, and 42.4% lived away from their family (35.2% in apartment and 7% in residence). Finally, the mean family income ranged between Canadian $40,000 and $49,999 CAN. On average, participants’ mothers had at least a high school diploma and fathers, a college degree.

Procedure

Research assistants took appointments with 3 college teachers and 2 university professors in order to ask students in their classes to partake in a study on vocational development. Research assistants explained to students the goals of the study and asked them to raise their hands if they wanted to participate. Those who accepted received a consent form and a questionnaire. Those who declined were asked to work quietly. No compensation was offered to students for their participation.

Measures

Vocational interests

The Activities section of the French Canadian version of the SDS was used to assess students’ vocational interests. The scale asks participants to indicate whether they would like to engage in each of the 66 proposed career-related activities. The original version includes 11 activities per dimension of the RIASEC. Examples of activities include paint (Artistic), teach (Social), and sell (Enterprising). In the original version, participants use a Yes−No response format. In order to increase the variability in participants’ answers (Hogan, 2007), we modified the original scale—with the approval of Psychological Assessment Resources, Inc.—to a 5-point Likert-type scale ranging from I really don’t like or wouldn’t like to do this activity (1) to I really like or would like to do this activity (5). Specifically, for research purposes it is better to use a 5-point scale instead of a Yes−No forced choice scale since this increases the variability of the scores, which can increase reliability estimates (Hogan, 2007). Using such a response format can also increase the probabilities of obtaining better fit indices in subsequent CFA analyses.

Statistical Analyses

IRT

IRT (see Embretson & Reise, 2000) is a type of nonlinear factor analysis and can be used to complement traditional scale development methods when selecting an optimal sample of items from a larger pool (Kristjansson, Breithaupt, & McDowell, 2001). One of the goals of IRT is to estimate the probability of endorsing an item as a function of a latent variable (θ). This probability is estimated by a curve (Hambleton, Swaminathan, & Rogers, 1991; Santor, Ramsay, & Zuroff, 1994). In our study, the level of interest for each RIASEC dimension represents the latent variable (θ). Several models of IRT exist, but with the use of Likert scales, the Graded-Response Model (GRM; Samejima, 1969, 1997) was deemed the most appropriate (Embretson & Reise, 2000; van der Linden & Hambleton, 1997). In this study, we used the Excel Item Response Theory assistant (EIRT) program (Germain, Valois, & Abdous, 2007) to compute the parameter estimates under the GRM.

GRM was first used to estimate which items best differentiated participants’ answers as a function of their respective interest level. First, we examined the item slope (a_i ) parameter, which indicates the location of the category response curves (CRCs). The higher the a_i , the more response categories should differentiate among interest levels. We also evaluated the b_ij parameters (one for each CRC) and their standard errors. The b_ij parameters represent a given interest level as a function of the maximal probability of choosing each response option. A good item will have b_ij parameters that match their respective response categories and a low error level. We also used item informativeness (I_i ) to assess the accuracy of the interest level estimation (Baker, 2001). When the amount of information is large, the latent trait level (interest) is more accurately estimated (Baker, 2001). Informativeness can be computed for each latent trait level and represented graphically as an item information curve (IIC).

Structural equation modeling (SEM)

We also used SEM (EQS program, version 6.1) to perform a CFA in order to select items. The adequacy of the CFA solution was assessed with different goodness-of-fit criteria, notably the chi-square. With large sample size, the chi-square is usually significant. For this reason, we also evaluated model goodness of fit with other fit indices such as the Comparative Fit Index (CFI), Non-Normed Fit Index (NNFI), Root Mean Square Error of Approximation (RMSEA), and the chi-square-to-degree of freedom ratio (χ²/df). The NNFI and CFI vary along a 0-to-1 continuum in which values greater than .90 are considered acceptable (Schumacker & Lomax, 1996). Browne and Cudeck (1993; also see Jöreskog & Sörbom, 1993) suggested that RMSEA values lower than .05 are indicative of a close fit and values up to .08 represent reasonable errors of approximation. There is no clear guideline about the value of the χ²/df, but usually a value lower than 3 reflects acceptable goodness of fit (Kline, 1998).

Missing data

A total of 700 participants completed all items of the questionnaire. Only 26 participants (3.58%) had missing values. We used the Little’s Missing Completely at Random (MCAR) test to verify whether the pattern of missing values was completely at random (MCAR). This test is based on a chi-square distribution with F degrees of freedom. If the test is nonsignificant, the pattern of missing values is MCAR. The Little’s MCAR test, χ² = 1569.079, df = 1533, p = .26, was nonsignificant, thereby justifying a listwise deletion of missing cases. Specifically, a listwise deletion of cases produces adequate parameters’ estimates under MCAR (see Peugh & Enders, 2004).

Results

Is the 66-item French Canadian version equivalent to the 66-item English version?

To test whether the 66-item French Canadian version was equivalent to the English version, as presented in the SDS technical manual (Holland et al., 1994), we compared reliability coefficients for both gender and correlations among RIASEC dimensions. Kuder-Richardson-20 coefficients (KR-20) were used with the American sample because the response format was dichotomous. In the present study, we used Cronbach’s coefficients because responses used a Likert-type format. For more information regarding these coefficients, see Kaplan and Saccuzzo (1993). As can be seen in Table 1, all reliability coefficients for English and French versions were acceptable (i.e., higher than .70; Nunally, 1978). To compare reliability coefficients, we use Feld’s test for independent samples, W = (1 − α₂/1 − α₁), where W can be interpreted as an F statistic with (N ₁ − 1), (N ₂ − 1) degrees of freedom; Feldt, 1969). Also, because we conducted several tests on the same data (6 tests for each gender), all p values were adjusted using a Bonferroni correction (Holm, 1979). Overall, we obtained three significant differences between reliability coefficients of the French version and those of the English one with those of the French version being highest.

Table 1.

Study 1: Comparisons of Reliability Coefficient Between the 66-item English and French Versions

Dimension	Men				Women
Dimension	KR-20 66-Item English^a	α 66-Item French^b	W	Stepdown Bonferroni p	KR-20 66-Item English^c	α 66-Item French^d	W	Stepdown Bonferroni p
Realistic	.86	.88	.857	.63	.83	.88	.706^*	.00
Investigative	.84	.89	.688	.02	.83	.87	.765^*	.01
Artistic	.82	.87	.722	.04	.79	.87	.619^*	.00
Social	.80	.84	.800	.28	.76	.81	.792^*	.03
Enterprising	.80	.88	.600	.00	.81	.85	.789^*	.03
Conventional	.88	.88	1.000	.99	.85	.86	.933	.80

Note: Kuder-Richardson-20 coefficients for the 66-item English version were reproduced by special permission of the Publisher, Psychological Assessment Resources, Inc., from the Self-Directed Search Form R Assessment Booklet by John L. Holland, Ph.D., Copyright 1970, 1977, 1985, 1990, 1994, 1995. Further reproduction is prohibited without permission from PAR, Inc.

^* p < .008; Two-tailed test.

^a N = 399.

^b N = 211.

^c N = 716.

^d N = 511.

Next, we compared correlations among the RIASEC dimensions of the French version to those presented in the SDS technical manual using a Z score difference and a Cohen’s q effect size (Cohen, 1977; Rossi, 1985). The Z score tests whether the difference between two correlation coefficients is significant, whereas the Cohen’s q effect size specifies the magnitude of this difference. Because we computed several tests (15 tests for each gender), we applied a Bonferroni correction to adjust the significance level (α/n; Shaffer, 1995). The adjusted significance level was therefore set at p < .003 (.05/15 = .003). Of the 30 correlations, 9 (30%) were significantly different (see Table 2) although the effect size of these differences where small for most of them (only 2 were moderate). No large effect size was observed. In addition, few of the significant differences were problematic in light of Holland’s theory. Specifically, most significant differences among correlations paralleled the organization of the six personality types on the hexagon where distant types of interest were more weakly correlated than proximal ones. However, two significant differences between correlations appeared problematic for the French version. Specifically, the S-E and S-C correlations for the men sample were, respectively, .36 and .22 in the English version (r = .05; p = .45), but only .05 and −.07 in the French version (r = −.07; p = .31). Nevertheless, we conclude that overall the 66-item French Canadian version is relatively equivalent to the English version in terms of reliability and correlations, thereby allowing the possibility to shorten the 66-item French version.

Table 2.

Study 1: Comparisons of Correlation among RIASEC Dimensions for the 66-item English and French Versions

Dimension	Men					Women
Dimension	66-Item English^a	66-Item French^b	Z Score	p	q Effect Size	66-Item English^c	66-Item French^d	Z Score	p	q Effect Size
R-I	.21	.35	1.99	.046	.163	.27	.31	0.86	.391	.054
R-A	.08	.37	4.04^*	.000	.323	.28	.47	4.37^*	.000	.175
R-S	−.02	.14	2.11	.035	.151	.06	−.02	−1.58	.116	.050
R-E	.07	.09	0.26	.792	.050	.13	.05	−1.64	.113	.050
R-C	.20	.17	−0.41	.684	.052	.22	.21	−0.20	.837	.000
I-A	.19	.20	0.14	.892	.000	.12	.15	0.60	.549	.051
I-S	.26	.37	1.60	.109	.056	.13	.25	2.45	.014	.155
I-E	.18	.14	−0.54	.591	.052	.13	.12	−0.20	1.000	.000
I-C	.26	.30	0.59	.569	.054	.21	.38	3.67^*	.000	.221
A-S	.32	.31	−0.15	.884	.000	.29	.21	−1.67	.094	.107
A-E	.13	−.05	−2.37	.018	.050	.22	.03	−3.80^*	.000	.153
A-C	−.01	−.03	−0.26	.793	.050	.02	−.14	−3.16^*	.001	.151
S-E	.36	.05	−4.28^*	.000	.374	.26	.12	−2.86	.004	.158
S-C	.22	−.07	−3.85^*	.000	.102	.04	−.07	−2.16	.031	.100
E-C	.47	.67	3.94^*	.000	.226	.36	.55	4.74^*	.000	.195

Note: Correlations for the 66-item English version were reproduced by special permission of the Publisher, Psychological Assessment Resources, Inc., from the Self-Directed Search Form R Assessment Booklet by John L. Holland, Ph.D., Copyright 1970, 1977, 1985, 1990, 1994, 1995. Further reproduction is prohibited without permission from PAR, Inc.

R = Realistic; I = Investigative; A = Artistic; S = Social; E = Enterprising; C = Conventional.

^a N = 1002.

^b N = 210.

^c N = 1600.

^d N = 511.

* p < .003.

Which items should be selected?

To determine which items to select, we used three parameters of the GRM, namely a_i , b_ij , and I_i , as well as factor loadings from a CFA. In the first step, we used the GRM parameters to choose items that best differentiated the participants in terms of interest level. The item slope (a_i ) parameter was first examined. As mentioned above, the higher the a_i , the better response categories differentiate the interest level. Thus, based on Baker (2001), we kept items with a_i > 1.35, which corresponds to a high differentiation level. We also used the b_ij parameters and their standard errors. We kept items with b_ij parameters having an adequate location on the interest continuum and the lowest standard error (<.30). Following a joint analysis of the a_i and b_ij parameters, we eliminated 23 items. Table 3 presents the items that were dropped in this first step. Next, we examined participants’ IICs for the 43 remaining items and kept, for each dimension, the 5 items that most accurately estimated interest levels. We thereby eliminated 13 additional items (see Table 4).

Table 3.

Study 1: a_i Parameters for the 66 Option Characteristic Curves and b_ij Parameters and Standard Errors of the 66 Category Response Curves

	a_i	b ₁ (SE)	b ₂ (SE)	b ₃ (SE)	b ₄ (SE)	b ₅ (SE)		a_i	b ₁ (SE)	b ₂ (SE)	b ₃ (SE)	b ₄ (SE)	b ₅ (SE)
Realistic							Social
Item 1	2.74	−0.27 (.04)	0.09 (.03)	0.79 (.08)	1.58 (.15)	2.05 (.20)	Item 1	2.03	−2.18 (.20)	−1.83 (.16)	−0.99 (.09)	0.12 (.04)	0.74 (.08)
Item 2	3.61	−0.13 (.03)	0.17 (.03)	0.69 (.07)	1.35 (.13)	1.79 (.18)	Item 2	1.25	−2.29 (.22)	−1.68 (.16)	−0.51 (.07)	1.01 (.11)	1.97 (.19)
Item 3	4.74	−0.26 (.04)	0.05 (.02)	0.60 (.06)	1.26 (.12)	1.66 (.17)	Item 3	0.72	−3.20 (.40)	−2.44 (.30)	−0.70 (.13)	1.72 (.23)	3.15 (.39)
Item 4	1.86	−1.11 (.11)	−0.66 (.07)	0.13 (.04)	1.13 (.10)	1.79 (.16)	Item 4	2.02	−2.63 (.26)	−2.28 (.22)	−1.55 (.15)	−0.58 (.07)	0.02 (.05)
Item 5	1.04	−0.27 (.09)	0.27 (.07)	1.38 (.16)	2.70 (.29)	3.43 (.38)	Item 5	1.40	−2.19 (.21)	−1.68 (.16)	−0.60 (.07)	0.72 (.08)	1.48 (.14)
Item 6	2.62	−0.24 (.04)	0.15 (.03)	0.83 (.08)	1.62 (.15)	2.10 (.21)	Item 6	0.97	−3.74 (.41)	−2.98 (.31)	−1.51 (.17)	0.14 (.08)	1.08 (.14)
Item 7	1.61	−0.37 (.06)	−0.01 (.05)	0.67 (.08)	1.51 (.14)	2.01 (.20)	Item 7	2.14	−1.90 (.17)	−1.59 (.14)	−0.96 (.09)	−0.07 (.04)	0.53 (.06)
Item 8	0.53	−3.85 (.62)	−2.87 (.47)	−1.16 (.24)	1.02 (.18)	2.48 (.38)	Item 8	0.60	−4.78 (.66)	−3.89 (.53)	−1.99 (.29)	0.46 (.13)	1.90 (.29)
Item 9	0.99	−1.36 (.17)	−0.68 (.11)	0.68 (.09)	2.33 (.24)	3.30 (.35)	Item 9	2.09	−1.80 (.17)	−1.53 (.14)	−0.97 (.09)	−0.12 (.04)	0.45 (.06)
Item 10	2.20	−0.79 (.08)	−0.39 (.05)	0.27 (.04)	1.05 (.09)	1.57 (.14)	Item 10	1.28	−3.54 (.37)	−2.94 (.29)	−1.79 (.18)	−0.25 (.06)	0.74 (.10)
Item 11	3.27	−0.20 (.04)	0.08 (.03)	0.54 (.06)	1.05 (.10)	1.37 (.13)	Item 11	0.56	−3.02 (.44)	−2.36 (.35)	−0.83 (.17)	1.09 (.20)	2.14 (.33)
Investigative						Enterprising
Item 1	1.18	−2.32 (.23)	−1.64 (.17)	−0.49 (.08)	0.97 (.11)	1.95 (.19)	Item 1	0.76	−6.09 (.80)	−5.17 (.64)	−2.92 (.36)	−0.27 (.10)	1.04 (.16)
Item 2	1.78	−1.01 (.10)	−0.60 (.07)	0.30 (.05)	1.39 (.12)	2.01 (.18)	Item 2	1.06	−1.56 (.17)	−0.89 (.11)	0.31 (.07)	1.93 (.20)	3.02 (.31)
Item 3	3.10	−0.24 (.04)	0.08 (.03)	0.67 (.07)	1.46 (.13)	2.00 (.20)	Item 3	2.52	−1.39 (.12)	−1.05 (.09)	−0.35 (.05)	0.58 (.06)	1.15 (.10)
Item 4	2.21	−1.24 (.11)	−0.81 (.08)	0.02 (.04)	0.92 (.08)	1.43 (.12)	Item 4	1.82	−2.15 (.20)	−1.84 (.17)	−1.03 (.10)	0.06 (.04)	0.66 (.07)
Item 5	1.97	−0.90 (.09)	−0.53 (.06)	0.20 (.04)	1.14 (.10)	1.73 (.16)	Item 5	1.56	−0.99 (.10)	−0.47 (.06)	0.47 (.06)	1.57 (.14)	2.23 (.21)
Item 6	3.32	−0.33 (.04)	0.05 (.03)	0.72 (.07)	1.54 (.14)	2.10 (.20)	Item 6	2.32	−0.96 (.09)	−0.59 (.06)	0.10 (.04)	0.95 (.08)	1.49 (.13)
Item 7	0.49	−6.15 (1.00)	−5.06 (.82)	−2.93 (.49)	0.04 (.14)	1.96 (.34)	Item 7	2.04	−2.17 (.20)	−1.76 (.15)	−0.87 (.08)	0.22 (.04)	0.82 (.08)
Item 8	1.33	−0.71 (.09)	−0.26 (.06)	0.57 (.08)	1.59 (.15)	2.22 (.22)	Item 8	1.43	−2.78 (.26)	−2.26 (.21)	−1.15 (.12)	0.27 (.06)	1.11 (.12)
Item 9	2.63	−0.25 (.04)	0.09 (.03)	0.75 (.07)	1.56 (.14)	2.06 (.20)	Item 9	2.34	−1.13 (.10)	−0.82 (.07)	−0.13 (.04)	0.74 (.07)	1.25 (.11)
Item 10	1.23	−0.52 (.08)	−0.01 (.06)	0.85 (.10)	1.88 (.19)	2.54 (.26)	Item 10	1.37	−2.73 (.26)	−2.32 (.22)	−1.33 (.13)	0.07 (.05)	0.91 (.10)
Item 11	1.66	−1.10 (.11)	−0.74 (.08)	−0.09 (.05)	0.89 (.09)	1.60 (.15)	Item 11	0.78	−0.05 (.10)	0.44 (.11)	1.54 (.21)	3.19 (.39)	4.24 (.52)
Artistic							Conventional
Item 1	1.74	−1.62 (.15)	−1.17 (.11)	−0.40 (.06)	0.43 (.06)	0.95 (.09)	Item 1	2.03	−0.00 (.05)	0.41 (.06)	1.18 (.12)	2.04 (.20)	2.55 (.26)
Item 2	1.36	−1.44 (.14)	−1.03 (.11)	−0.23 (.06)	0.79 (.09)	1.41 (.14)	Item 2	0.49	−3.22 (.54)	−1.83 (.33)	0.66 (.17)	4.02 (.63)	6.27 (.99)
Item 3	1.10	−1.93 (.20)	−1.43 (.15)	−0.35 (.07)	1.04 (.12)	1.86 (.19)	Item 3	3.74	−0.60 (.06)	−0.26 (.03)	0.30 (.04)	0.91 (.08)	1.30 (.12)
Item 4	1.47	−1.16 (.12)	−0.71 (.08)	0.05 (.05)	0.90 (.09)	1.43 (.14)	Item 4	1.07	−0.90 (.12)	−0.18 (.07)	1.28 (.14)	2.95 (.30)	3.89 (.42)
Item 5	1.41	−1.54 (.15)	−1.19 (.12)	−0.54 (.07)	0.41 (.06)	1.05 (.11)	Item 5	3.28	−0.63 (.06)	−0.26 (.04)	0.37 (.04)	1.11 (.10)	1.57 (.14)
Item 6	0.84	−3.36 (.38)	−2.63 (.30)	−1.11 (.15)	0.77 (.12)	1.85 (.22)	Item 6	1.32	−1.27 (.13)	−0.76 (.09)	0.30 (.06)	1.68 (.16)	2.52 (.24)
Item 7	2.03	−1.71 (.16)	−1.37 (.12)	−0.76 (.08)	−0.00 (.04)	0.49 (.06)	Item 7	1.40	−1.76 (.17)	−1.32 (.13)	−0.43 (.07)	0.72 (.08)	1.44 (.14)
Item 8	1.68	−1.06 (.10)	−0.66 (.07)	0.08 (.05)	0.89 (.09)	1.36 (.13)	Item 8	4.05	−0.58 (.06)	−0.24 (.03)	0.37 (.04)	1.11 (.10)	1.57 (.15)
Item 9	1.56	−0.80 (.09)	−0.42 (.06)	0.29 (.06)	1.10 (.11)	1.58 (.15)	Item 9	2.69	−0.38 (.05)	−0.03 (.03)	0.64 (.06)	1.40 (.12)	1.85 (.17)
Item 10	2.84	−1.14 (.10)	−0.85 (.08)	−0.33 (.04)	0.28 (.04)	0.66 (.06)	Item 10	0.63	−4.31 (.59)	−3.30 (.46)	−1.41 (.23)	1.16 (.18)	2.85 (.38)
Item 11	1.64	−0.82 (.09)	−0.41 (.06)	0.37 (.06)	1.18 (.11)	1.62 (.15)	Item 11	1.29	−0.43 (.08)	0.08 (.06)	1.12 (.12)	2.53 (.25)	3.41 (.36)

Note. Crossed-out items are the ones deleted from the entire pool of items.

Table 4.

Study 1: Maximal Informativeness of the 43 IIC

Items	I_i ^max	Items	I_i ^max
Realistic		Social
Item 1	1.60	Item 1	0.85
Item 2	2.65	Item 4	0.81
Item 3	5.25	Item 5	0.37
Item 4	0.72	Item 7	0.93
Item 6	1.42	Item 9	0.89
Item 7	0.47
Item 10	0.90
Item 11	2.20
Investigative		Enterprising
Item 2	0.63	Item 3	1.40
Item 3	2.20	Item 4	0.47
Item 4	1.00	Item 5	0.65
Item 5	0.79	Item 6	1.10
Item 6	2.51	Item 7	0.85
Item 9	1.41	Item 8	0.42
Item 11	0.55	Item 9	1.10
		Item 10	0.38
Artistic		Conventional
Item 1	0.46	Item 1	0.80
Item 2	0.35	Item 3	3.00
Item 4	0.38	Item 5	2.30
Item 5	0.37	Item 7	0.37
Item 7	0.80	Item 8	3.10
Item 8	0.51	Item 9	1.50
Item 9	0.56
Item 10	1.52
Item 11	0.47

Note. Because item reproduction is prohibited without permission from PAR Inc., the labels of the items cannot be written. Nevertheless, their numbers are presented in the order in which they appear in the Activities section of the Assessment booklet, which can be purchased from PAR Inc. Crossed-out items are the ones deleted from the entire pool of items.

In the second step, we used a CFA to examine the factor loadings of the 30 remaining items, Satorra-Bentler (S-B) χ² = 1 578.43; df = 391; S-B χ²/df = 4.04; NNFI = .87; CFI = .89; RMSEA = .07. The low values of these fit indices suggest that some items were problematic. In line with our goal to create a 24-item scale, we selected the 4 most optimal items to assess each dimension. The 6 items with the lowest factor loadings were thus eliminated (see Table 5). In sum, based on the parameters a_i , b_ij, and I_i and the factor loadings, we selected 24 items that best captured the s6 vocational interests’ dimensions.

Table 5.

Study 1: Factor Loadings of the 30 Items From the CFA Solution

Item	Factor Loading	Item	Factor Loading
Realistic		Social
Item 1	.754	Item 1	.705
Item 2	.868	Item 4	.698
Item 3	.897	Item 5	.581
Item 6	.695	Item 7	.763
Item 11	.823	Item 9	.727
Investigative		Enterprising
Item 3	.840	Item 3	.813
Item 4	.644	Item 5	.607
Item 5	.596	Item 6	.865
Item 6	.874	Item 7	.527
Item 9	.755	Item 9	.740
Artistic		Conventional
Item 7	.530	Item 1	.676
Item 8	.743	Item 3	.861
Item 9	.725	Item 5	.814
Item 10	.727	Item 8	.879
Item 11	.635	Item 9	.779

Note. Because item reproduction is prohibited without permission from PAR Inc., the labels of the items cannot be written. Crossed-out items are the ones deleted from the entire pool of items.

Summary

Results showed that the 66-item French Canadian version of the Activities section of the SDS is relatively equivalent to the original English version in terms of reliability and correlations. In addition, using IRT and CFA analyses, we selected 24 items. However, we have yet to determine whether the factor structure of these selected items is adequate, whether it is invariant across gender, and whether the correlations among the subscales are in line with Holland’ theory. These will be the goals of Study 2.

Study 2

Method

Participants

A total of 338 French Canadian young adults (95 men; 243 women) agreed to partake in this study. Of these participants, 252 attended college, 58 attended university, and 28 had a full-time job. Participants’ mean age was 19.82 years, SD = 2.91, range = (16−44), with 5 participants being over 30). Among college students, 24.2% were in humanities or social sciences, 19.8% in natural sciences, 7.1% in arts, and 48.8% in other domains. Among university students, 27.5% were in a teacher education program, 17.2% in administration, 12.3% in engineering, 10.3% in a business program, and 32.7% were in other programs. A total of 6.1% of college students had academic achievement scores above 90%, 82% had scores between 70% and 80%, and 11.9% had scores lower than 70%. For university students, 6.9% of them had academic achievement scores above 90%, 81% had scores between 70% and 80%, and 12.1% had scores below 70%. For the 28 participants who worked full time, 82.1% had a college degree, 14.3% a high school diploma, and 3.6% a bachelor’s degree.

A total of 52.2% of the students lived with both parents, 19.9% lived away from their family (18.1% in apartment and 1.8% in residence), and 27.9% lived with one parent. Overall, 96.7% were born in the province of Quebec, 1.5% in another Canadian province, and 1.8% in another country. Most of them (98.2%) spoke French as their first language (0.9% English and 0.9% another language). Parents’ average annual income was between Canadian $40,000 and $49,999 CAN and the majority of mothers (25.7%) and fathers (21.3%) had at least a college degree.

Procedure

Participants for this study were a subsample from a larger study on vocational development. All participants from this larger sample (n = 975) were contacted by phone and asked to participate in the present study, which consisted in completing an online survey that took approximately 20 minutes to fill out. They were told that 25 pairs of movie tickets would be drawn at random among those who had completed the survey. Upon participants’ acceptance, we gave them an ID and a password to access the online survey. Additional phone calls were made to remind participants to complete the survey. The 338 participants who accepted to participate in the study had no missing values (they answered all items of the survey). This could be explained by the fact that participants were unable to submit their answers if they had not completed all items.

Statistical Analyses

As in Study 1, we performed a CFA but, here, we tested the invariance of the CFA solution across gender. We first computed separate covariance matrices for men and women. The minimal condition of factorial invariance is the invariance of the factor loadings (Model 1). Thus, separate tests were conducted to test the invariance of the factor loadings (Model 2) as well as the invariance of variances and covariances (Model 3; Little, Preacher, Selig, & Card, 2007). We used the chi-square difference test at p < .01 to verify whether the factor structure was invariant across gender. As recommended by Cheung and Rensvold (2002), we used the .01 difference in CFI among invariance models. Specifically, if the difference in CFI between two models was lower than .01, we concluded that both models were equivalent.

Measures

Vocational interests

The 24 items selected in Study 1 were used to assess vocational interests. Cronbach’s αs were adequate, α_R =.93; α_I =.87; α_A =.79; α_S =.85; α_E =.92; α_C =.92, thereby providing support for the reliability of each dimension.

Results

Do the selected items have adequate construct validity?

To determine whether the selected items presented adequate construct validity, we first examined whether the factorial structure underlying the RIASEC model was reproduced by the data. To do so, we performed a CFA on the 24 selected items and examined the goodness-of-fit criteria. Results showed adequate fit indices, S-B χ² = 495.55; df = 236; S-B χ²/df = 2.10; NNFI = .94; CFI = .95; RMSEA = .06. In addition, all factor loadings were above .50 (see Table 6).

Table 6.

Study 2: Factor Loadings of the 24 Items From the CFA Solution

Item	Factor Loading	Item	Factor Loading
Realistic		Social
Item 1	.851	Item 1	.701
Item 2	.821	Item 2	.788
Item 3	.957	Item 3	.832
Item 4	.799	Item 4	.748
Investigative		Enterprising
Item 1	.907	Item 1	.844
Item 2	.742	Item 2	.696
Item 3	.884	Item 3	.923
Item 4	.668	Item 4	.779
Artistic		Conventional
Item 1	.846	Item 1	.883
Item 2	.808	Item 2	.785
Item 3	.616	Item 3	.872
Item 4	.509	Item 4	.892

Note. Because item reproduction is prohibited without permission from PAR Inc., the labels of the items cannot be written.

Are the selected items invariant across gender?

We tested the selected items for factorial invariance across gender. We computed three models. In the least restrictive model (Model 1), no parameter was constrained to be equal across gender. This model provided a relatively good approximation of the data, S-B χ² = 836.22; df = 472; χ²/df = 1.77; CFI = .94; NNFI = .93; RMSEA = .06. In Model 2, factor loadings were constrained to be invariant across gender, S-B χ² = 855.43; df = 490; χ²/df = 1.75; CFI = .94; NNFI = .93; RMSEA = .06. The chi-square difference between these two models was not significant, χ² _diff = 22.04; df _diff = 18; p > .01. In Model 3, variances and covariances were constrained to be invariant across gender, S-B χ² = 881.40; df = 505; χ²/df = 1.75; CFI = .94; NNFI = .93; RMSEA = .06. The chi-square difference between models 2 and 3 was not significant, χ² _diff = 25.74; df _diff = 15; p > .01. In addition, there were no differences among the CFI of the three models, ΔCFI = .00. Based on these results, we can conclude that factor loadings, variances, and covariances were invariant across gender.

Are correlations among interest dimensions of the selected items equivalent to those of the original English version?

We compared correlations among the RIASEC dimensions for the 24 selected items to those of the 66-item English version presented in the SDS Technical Manual. Because correlations are presented in the technical manual as a function of gender, we compared correlations for each gender separately, even though we found no differences in correlational patterns in the invariance analyses. As in Study 1, we used a Z score difference and a Cohen’s q and we adjusted the α level via a Bonferroni correction. As we can observe in Table 7, five correlations reached significance (one for men and four for women). Based on Cohen’s criteria for estimating effect size (Cohen, 1977), we can qualify these differences as varying from small to moderate. No effect size was observed. Some differences were noteworthy. For men, the correlation between I-S was .26 for the 66-item English version, but −.24 (p = .02) for the 24-item French version. For women, correlations between I-C, A-E, S-E, and E-C were, respectively, .21, .22, .26, and .36 for the 66-item English version but .40 (p = .00), −.01 (p = .93), −.01 (p = .93), and .55 (p = .00) for the 24 selected items.

Table 7.

Study 2: Comparisons of Correlation among RIASEC Dimensions for the Original 66-item English and the 24 selected items from the French-Canadian Version

	Men					Women
Dimension	66-item English^a	24-item French^b	Z score	p	q effect size	66-item English^c	24-item French^d	Z score	p	q effect size
R - I	.21	.45	2.49	.013	.282	.27	.34	1.12	.265	.056
R - A	.08	−.10	−1.66	.098	.000	.28	.17	−1.68	.094	.158
R - S	−.02	−.26	−2.26	.024	.255	.06	.08	0.29	.772	.000
R - E	.07	.09	0.18	.853	.000	.13	.07	−0.88	.381	.050
R - C	.20	.27	0.68	.496	.053	.22	.15	−1.05	.295	.052
I - A	.19	−.08	−2.50	.012	.102	.12	.07	−0.73	.466	.050
I - S	.26	−.24	−4.69^*	.000	.053	.13	.04	−1.31	.190	.050
I - E	.18	−.05	−2.13	.033	.153	.13	.11	−0.29	.769	.000
I - C	.26	.19	−0.68	.498	.053	.21	.40	3.04^*	.002	.221
A - S	.32	.28	−0.40	.687	.000	.29	.23	−0.93	.352	.054
A - E	.13	−.02	−1.38	.167	.151	.22	−.01	−3.38^*	.001	.203
A - C	−.01	−.10	−0.83	.407	.100	.02	−.13	−2.18	.029	.151
S - E	.36	.10	−2.54	.011	.265	.26	−.01	−3.99^*	.000	.255
S - C	.22	−.03	−2.33	.020	.153	.04	−.05	−1.30	.193	.050
E - C	.47	.62	1.97	.049	.208	.36	.55	3.49^*	.000	.195

Note. Correlations for the 66-item English version were reproduced by special permission of the Publisher, Psychological Assessment Resources, Inc., 16204 North Florida Avenue, Lutz, Florida 33549, from the Self-Directed Search Form R Assessment Booklet by John L. Holland, Ph.D., Copyright 1970, 1977, 1985, 1990, 1994, 1995. Further reproduction is prohibited without permission from PAR, Inc.

R = Realistic; I = Investigative; A = Artistic; S = Social; E = Enterprising; C = Conventional.

^* p < .003.

^aN = 1002.

^bN = 95.

^cN = 1600.

^dN = 243.

Summary

The findings of Study 2 showed that the selected pool of items of the Activities section of the SDS is adequate and that its factor structure was invariant across gender. In addition, correlations among subscales were in the expected direction, except for 5 differences between the original 66-item English version and the 24 selected items (one for men and four for women). One may argue that these differences originate from the 24 selected items. However, it should be noted that the differences for I-C, A-E, and E-C for the women sample also appear when we compared the 66-item of both French and English versions (see Study 1). These unexpected findings might consequently have more to do with the original, longer version. Future studies are needed to replicate those differences and to explain them.

General Discussion

The main goal of the present research was to select a restricted pool of items from the French Canadian version of the Self-Directed Search—Activities section (Holland, 1991). In Study 1, results showed that the 66-item French Canadian version was equivalent to the original English version regarding reliability and correlations among the 6 types of interests. Using IRT and CFA, we selected 24 items from the 66-item version. In Study 2, results showed that the selected items had adequate factorial validity and that they were invariant across gender. In addition, correlations among subscales were in the expected direction, although some differences emerged between the original 66-item English version and the 24 selected items. The implications of these results are discussed below.

Implications for Research, Theories, and Career Counseling

The findings we obtained lead to a number of implications for research, theories, and career counseling. First, by testing the equivalence of the 66-item French Canadian version of the Activities section to the 66-item English version, the results of Study 1 provided further support for the French version of this SDS subscale. To the best of our knowledge, this is the first study that directly compared the psychometric properties of these two versions. It thus provides some support for using the SDS French Canadian version with college and university students. However, as we noted, some differences among correlations exist between both versions, specifically for men regarding social, enterprising, and conventional types of vocational interests. That is, in the American men sample, the correlation between social and enterprising dimensions was moderate and positive whereas, in the French Canadian men sample, this correlation was quite low. Moreover, in the American men sample, the correlation between social and conventional was positive, but negative in the French Canadian men sample. Because these types of interests are adjacent on the hexagon, one would expect higher, positive correlations among these dimensions for the French Canadian sample. However, these differences in correlation were not significant when we compared American women and French Canadian women samples. At this point, we cannot explain such differences in the French Canadian male sample. Are they due to cultural differences (i.e., occupational activities in certain countries might be different from those in the United States, where the original scale was developed), sample characteristics, or gender differences in interpretation of the items? Future research is needed to reproduce these unexpected correlations and explain them. More research is also required to replicate the RIASEC structure with Canadian samples.

Second, results from IRT and CFA allowed us to select 24 items. As mentioned earlier, selecting a reduced pool of items when researchers face time constraints has the advantages of diminishing item redundancy and reducing participants’ fatigue, frustration, and boredom (Robins et al., 2001). We must, however, keep in mind that the selected items focus on vocational interests but not other aspects of the vocational development such as competencies, occupation, or self-estimates. In addition, the 24 selected items might be different from those that would be selected in another language (i.e., English, German). Although the strength of this research was that the 24 items were selected through strict statistical analysis, it is possible that using different samples in different countries might yield different choices of items. This possibility need to be investigated in future studies.

Third, results from Study 2 showed that the selected items were reliable and had an adequate factorial structure that it is invariant across gender. These results parallel those of Nagy, Trautwein, and Lüdtke (2010), which demonstrated in a sample of German students that the structure of vocational interests is invariant across gender. However, before using those selected items for research purposes, it would be important to test it on various populations including high school students, young adults, and adults on the job market. Although we found a psychometrically sound reduced pool of items in the SDS Activities section, we encourage researchers and practitioners to use the complete version when they have sufficient time.

Fourth, most correlations for the selected items were in line with those obtained with the original 66-item English version. It thus appears that having fewer items did not affect the direction or magnitude of correlations among interests’ dimensions. The pattern of the correlations was thus in line with Holland’s model in which interest types can be represented in a hexagonal fashion where the distance between each type represents the relative magnitude of the correlations. That is, correlations between adjacent types are greater than those between all nonadjacent types (e.g., r _R-I > r _R-S). However, as it was the case with the 66-item French Canadian version, some correlations with the 24 selected items appeared problematic, especially, in Study 2, those involving the enterprising dimension for the women sample. Future research is thus needed on this issue.

Limitations

Despite the innovative nature of our findings, we need to consider some limitations when interpreting them. First, we did not verify extensively whether the SDS 66-item French Canadian version was equivalent to the English one. Conducting invariance analyses between English and French Canadian versions on factor loadings, variances, covariances, and uniquenesses appeared warranted. We were unable to conduct such an analysis because in the SDS English technical manual covariances among items were not reported.

Second, although we concluded that the psychometric qualities of the selected 24 items were adequate, other tests are needed to further test the psychometric properties of these selected items. For example, we did not test the scale’s reliability or construct validity over time. One advantage of using multiple time measurements would be to compute a multitrait−multimethod (MTMM) matrices to assess the scale’s convergent and divergent validities (see Campbell & Fiske, 1959; Guay, 2005; Marsh & Grayson, 1995 for examples). In addition, it could be useful to test the validity of the selected items using more sophisticated method of analyses such as the Randomization Test of Hypothesized Order Relations ([RTOR] see Nagy et al., 2010), which would test more fully the relations among the 6 dimensions as well as to test its construct validity with the 5-factor model of personality.

Finally, the samples used in Studies 1 and 2 were not representative of the entire college and university student populations. Our findings should be replicated in studies using larger, more representative samples that include students from ethnic minorities as well as those coming from diverse socioeconomic backgrounds.

Conclusion

Although there need to be additional tests of the validity of these 24 items selected from the Activities section of the SDS, the present study nevertheless underscores their psychometric properties. Thus, researchers working with French Canadian students could use these items to test their hypotheses concerning the determinants and consequences of career interests, when they do not have the required time to administer the 66-item version.

Footnotes

The Self-Directed Search was modified and reproduced with special permission of the Publisher, Psychological Assessment Resources, Inc., 16204 North Florida Avenue, Lutz, Florida, 33549, from the Self-Directed Search Form R Assessment Booklet by John L. Holland, PhD, copyright 1970, 1977, 1985, 1990, 1994, 1995. Further reproduction is prohibited without permission from PAR, Inc.

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

The author(s) disclosed receipt of the following financial support for the research and/or authorship of this article: SSHRC and by the financial support of the Canada Research Chair on Motivation and Academic Success.

References

Baker

F. B.

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

Browne

M. W.

Cudeck

(1993). Alternative ways of assessing model fit. In Bollen

Stine

(Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage.

Campbell

D. T.

Fiske

D. W.

(1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81–105.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255.

Cohen

(1977). Statistical power analysis for the behavioral sciences. New York, NY: Academic Press.

Deal

J. E.

Halverson

C. F.

Jr. Martin

R. P.

Victor

Baker

(2007). The Inventory of Children’s Individual Differences: Development and Validation of a Short Version. Journal of Personality Assessment, 89, 162–166.

Embretson

S. E.

Reise

S. P.

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

Feldt

L. S.

(1969). A test of the hypothesis that Cronbach’s alpha or Kuder-Richardson reliability coefficient twenty is the same for two tests. Psychometrika, 34, 363–373.

Gati

Asher

(2001). The PIC model for career decision making: Prescreening, in-depth exploration, and choice. In Leong

E. C.

Barak

(Eds.), Contemporary models in vocational psychology (pp. 7–54). Mahwah, NJ: Lawrence Erlbaum.

10.

Germain

Valois

Abdous

(2007). EIRT: module Excel pour effectuer des estimations de la Théorie de la réponse aux items. Retrieved April 1, 2009, from http://sourceforge.net/project/showfiles.php?group_id=164538&package_id=186544

11.

Glidden-Tracey

Greenwood

A. K.

(1997). A validation study of the Spanish Self-Directed Search using back-translation procedures. Journal of Career Assessment, 5, 105–113.

12.

Gosling

S. D.

Rentfrow

P. J.

Swann

W. B.

Jr. (2003). A very brief measure of the Big-Five personality domains. Journal of Research in Personality, 37, 504–528.

13.

Graham

J. W.

(2009). Missing data analysis: Making it work in the real world. Annual Review of Psychology, 60, 549–576.

14.

Guay

(2005). Motivations underlying career decision-making activities: The Career Decision-Making Autonomy Scale (CDMAS). Journal of Career Assessment, 13, 77–97.

15.

Guay

Ratelle

C. F.

Senécal

Larose

Deschênes

(2006). Distinguishing developmental from chronic career indecision: Self-efficacy, autonomy, and social support. Journal of Career Assessment, 14, 235–251.

16.

Guglielmi

Fraccaroli

Pombeni

M. K.

(2004). Les intérêts professionnels selon le modèle hexagonal de Holland, structures et différences de genre. Orientation scolaire et professionnelle, 33, 409–427.

17.

Hambleton

R. K.

Swaminathan

Rogers

H. J.

(1991). Fundamentals of item response theory. Thousand Oaks, CA: Sage.

18.

Hogan

T. P.

(2007). Psychological testing: a practical introduction. 2nd ed. Hoboken, NJ: John Wiley.

19.

Holland

J. L.

(1997). Making vocational choices: A theory of vocational personalities and work environments. 3rd ed. Odessa, FL: Psychological Assessment Resources.

20.

Holland

J. L.

(1991). L’orientation par soi-même: édition canadienne-française. Odessa, FL: Psychological Assessment Resources.

21.

Holland

J. L.

(1992). Making vocational choices: A theory of vocational personalities and work environment. 2nd ed. Odessa, FL: Psychological Assessment Resources.

22.

Holland

J. L.

Fritzsche

B. A.

Powell

A. B.

(1994). Self-directed search: Technical manual. Odessa, FL: Psychological Assessment Resources.

23.

Holm

(1979). A simple sequentially rejective Bonferroni test procedure. Scandinavian Journal of Statistic, 6, 65–70.

24.

Hood

A. B.

Johnson

R. W.

(2007). Assessment in Counseling: A guide to the use of psychological assessment procedures. Alexandria, VA: American Counseling Association.

25.

Jöreskog

K. G.

Sörbom

(1993). LISREL 8: Structural equation modelling with the SIMPLIS command language. Chicago, IL: Scientific Software International.

26.

Kapes

J. T.

Mastie

M. M.

Whitfield

E. A.

(1994). A counselor’s guide to career assessment instrument. Alexandria, VA: National Career Development Association.

27.

Kaplan

R. M.

Saccuzzo

D. R.

(1993). Psychological testing: Principles, applications, and issues (3rd ed.). Belmont, CA: Thomson Brooks/Cole Publishing.

28.

Khan

S. B.

Alvi

S. A.

(1991). The structure of Holland’s typology: A study in a Non-Western Culture. Journal of Cross-Cultural Psychology, 22, 283–292.

29.

Kline

R. B.

(1998). Principles and practice of structural equation modeling. New York, NY: Guilford Press.

30.

Kristjansson

Breithaupt

McDowell

(2001). Development and validation of an indicator of support for community-residing older Canadians. International Psychogeriatrics, 13, 125–135.

31.

Lent

R. W.

Brown

S. D.

Hackett

(2002). Social cognitive career theory. In Brown

Ass. (Eds.), Career choice & development (4th ed., pp. 255–311). San Francisco, CA: Jossey-Bass.

32.

Little

T. D.

Preacher

K. J.

Selig

J. P.

Card

N. D.

(2007). New developments in latent variable panel analyses of longitudinal data. International Journal of Behavioral Development, 31, 357–365.

33.

Marsh

H. W.

Grayson

(1995). Latent variable of multitrait-multimethod data. In Hoyle

R. H.

(Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 177–198). Thousand Oaks, CA: Sage.

34.

Marsh

H. W.

Hau

K.-T.

Balla

J. R.

Grayson

(1998). Is more ever too much? The number of indicators per factor in confirmatory factor analysis. Multivariate Behavioral Research, 33, 181–230.

35.

Muck

P. M.

(2005). Review of the German adaptation and enhancement of Holland’s Self-Directed Search. Zeitschrift für Personal Psychologie, 4, 39–46.

36.

Muck

P. M.

Hell

Gosling

S. D.

(2007). Construct validation of a short five-factor model instrument. European Journal of Psychological Assessment, 23, 166–175.

37.

Nagy

Trautwein

Lüdtke

(2010). The structure of vocational interests in Germany: Different methodologies, different conclusions. Journal of Vocational Behavior, 76, 153–169.

38.

Nunally

J. C.

(1978). Psychometric theory (2nd ed.). San Francisco, CA: Jossey-Bass.

39.

Oosterveld

(1994). Confirmatory factor analysis of the self-directed search test, a multitrait-multimethod approach. Personality and Individual Differences, 17, 565–569.

40.

Osipow

S. H.

Fitzgerald

L. F.

(1996). Theories of career development. Boston, MA: Allyn & Bacon Editions.

41.

Peugh

J. L.

Enders

C. K.

(2004). Missing data in educational research: A review of reporting practices and suggestions for improvement. Review of Educational Research, 74, 525–556.

42.

Rammstedt

John

O. P.

(2007). Measuring personality in one minute or less: A 10-item short version of the Big Five Inventory in English and German. Journal of Research in Personality, 41, 203–212.

43.

Robins

R. W.

Hendin

H. M.

Trzesniewski

K. H.

(2001). Measuring global self-esteem: Construct validation of a single-item measure and the Rosenberg Self-Esteem Scale. Personality and Social Psychology Bulletin, 27, 151–161.

44.

Rossi

J. S.

(1985). Tables of effect size for Z score tests of differences between proportions and between correlation coefficients. Educational and Psychological Measurement, 45, 737–743.

45.

Sabourin

Valois

Lussier

(2005). Development and validation of a brief version of the dyadic adjustment scale with nonparametric item analysis model. Psychological Assessment, 17, 15–27.

46.

Samejima

(1969). Estimation of ability using a response pattern of graded scores. Psychometrika Monograph Supplement, 34, 1–97.

47.

Samejima

(1997). Graded Response Model. In van der Linden

W. J.

Hambleton

R. K.

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

48.

Santor

D. A.

Ramsay

J. O.

Zuroff

D. C.

(1994). Nonparametric item analyses of the Beck Depression Inventory: Evaluating gender item bias and response option weights. Psychological Assessment, 6, 255–270.

49.

Schumacker

R. E.

Lomax

R. G.

(1996). A beginner’s guide to structural equation modeling. Hillsdale, NJ: Lawrence Erlbaum.

50.

Shaffer

J. P.

(1995). Multiple hypothesis testing. Annual Review of Psychology, 46, 561–584.

51.

Spokane

A. R.

Holland

J. L.

(1995). The self-directed search: A family of self-guided career interventions. Journal of Career Assessment, 3, 373–390.

52.

Turner

S. L.

Conkel

J. L.

Starkey

Landgraf

Lapan

R. T.

Siewert

J. J.

Ju-Ping

(2008). Gender differences in Holland vocational personality types: Implications for school counselors. Professional School Counseling, 11, 317–326.

53.

van der Linden

W. J.

Hambleton

R. K.

(1997). Handbook of modern item response theory. New York, NY: Springer-Verlag.

54.

Vondracek

F. W.

Porfeli

E. J.

(2003). The world of work and careers. In Adams

G. R.

Berzonsky

M. D.

(Eds.), Blackwell handbook of adolescence (pp. 109–128). Malden, MA: Blackwell.

55.

Yang

Lance

C. E.

Hui

H. C.

(2006). Psychometric properties of the Chinese self-directed search (1994 edition). Journal of Vocational Behavior, 68, 560–576.