Early Coursework and College Experience Predictors of Persistence at a Hispanic-Serving Institution

Abstract

Spanish

Utilizing survival analysis, we examined two key first year courses, mathematics and English, and tested whether they were predictive for long-term student success as measured by persistence to graduation at a Hispanic-serving research university. While first math and first English courses were significant, SAT score was not for Hispanic and Black students. High school grade point average (GPA) was significant for Hispanic students but not for White or Black students.

Keywords

math persistence Hispanic-serving Institution

Introduction

While there have been both qualitative and quantitative studies of Hispanic-serving institutions (HSI), most authors suggest further research is needed, particularly longitudinal studies (Arana, Castañeda-Sound, Blanchard, & Aguilar, 2011; Dayton, Gonzalez-Vasquez, Martinez, & Plum, 2004; Maestas, Vaquera, & Muñoz Zehr, 2007; Santiago, Andrade, & Brown, 2004). This study adds to this growing body of literature by examining key academic factors in the persistence to graduation of a diverse group of students at a HSI.

Several researchers have suggested that academic integration is more important than social integration to student success particularly for Latino students (Cabrera, Castañeda, Nora, & Hengstler, 1992; Nora & Cabrera, 1996; Torres, 2006). Santiago et al. (2004) recommend identifying gateway courses that can identify students who are struggling and be used as early measures of Latino student success at HSIs. Following this recommendation, we examined two key first year courses, mathematics and English, and tested whether they were predictive for long-term student success as measured by persistence to graduation. Utilizing survival analysis, we asked the following research question: Does early course performance in key classes (math and English composition) or other college academic experience variables predict persistence for a diverse group of students at a HSI?

Literature Review

While Tinto’s (1993) theory of academic and social integration has been dominant, tests showed that for Latino students, social integration was not significantly related to retention while academic integration was (Cabrera et al., 1992; Torres, 2006). An exception was Maestas et al.’s (2007) study of sense of belonging at a HSI where there were significant variables for both academic and social integration. Recently, alternative models of student retention (Nora, 2004; Perna & Thomas, 2008; Rendon, 1994; Torres, 2006) offer options that better account for important factors among Latino students. Here we review the prior research on HSIs looking at academic variables but briefly reviewing other studies to illustrate the need for several control variables such as precollege academic preparation measured by grade point average (GPA) and SAT and postsecondary experiences measured by first-year grades, academic major, financial aid, and enrollment intensity. While not controls in our model, a review of the literature would be incomplete without looking at faculty–student interaction and family encouragement.

Pre-College Academic Preparation

Pre-college academic preparation as measured by high school grades has been associated with college persistence in a number of studies (Bordes-Edgar, Arredondo, Kurpius, & Rund, 2011; Perna, 2006; Tinto, 1993). Yet, some have questioned the importance of this relationship for Latino and other minority students (Nora & Cabrera, 1996; Torres, 2006). Based on her qualitative interviews, Torres concluded Latino students often do not consider their K-12 academic performance illustrative of their potential. She concluded college experiences were more salient in the experience of Latino student than precollege measures. Likewise, in their analysis of persistence, Nora and Cabrera did not find a direct effect for precollege academic ability measures for Latino students.

Standardized admissions tests

The predictive power of standardized test scores is also unresolved for Latino and Black students in existing research (Arbona & Nora, 2007; Bordes-Edgar et al., 2011; Mattern, Patterson, Shaw, Kobrin, & Barbuti, 2008; Sackett, Kuncel, Arneson, Cooper, & Waters, 2009; Shen et al., 2012). The predictive power of SAT scores on first-year grades has been well established, and the general validity of test scores has research support (Mattern et al., 2008; Sackett et al., 2009). Yet, the validity of test scores for various subpopulations and the relationship with persistence has been questioned (Shen et al., 2012). Bordes-Edgar and colleagues found the SAT was not predictive of first-year grades or persistence in a longitudinal analysis of Latino students. They questioned the appropriateness of using the SAT with some populations. Arbona and Nora also found no significant relationship between test scores and persistence among Latino students. Shen and colleagues found SAT scores were less valid at larger schools, schools with more low-income students, and schools with a higher percentage of disadvantaged minority students, typical descriptors of HSIs.

Postsecondary Academic Experience

Measures of postsecondary academic experiences can include first-year grades, enrollment intensity (full-time/part-time), course performance, and participation in academic support services. Academic support and tutoring at HSIs have been positively associated with sense of belonging (Maestas et al., 2007) and student persistence (Otero, Rivas, & Rivera, 2007). While not specifically examined at HSIs, the research evidence from multiple studies suggests that attending full-time rather than part-time is associated with higher graduation rates (Arbona & Nora, 2007; St. John & Musoba, 2010).

Faculty and staff interaction are positively associated with student success. When faculty showed interest in students, HSI students had a better sense of belonging at the institution (Maestas et al., 2007). Torres (2006), studying Latino students at urban commuter campuses, establish a direct effect between meeting with faculty outside of class and students’ commitment to the institution.

First-year grades

Although not specifically studied at HSIs, studies of Latino students show college grades consistently predicted persistence (Arbona & Nora, 2007; Cabrera, Nora, & Castañeda, 1993; Nora, 2004; Nora & Cabrera, 1996; Otero et al., 2007; Torres, 2006). Nora and Cabrera showed that first-year grades were three times more important for Latino and African American students than White students in predicting persistence.

Academic major

Research on college major has been somewhat contradictory for Latino students. Traditionally, research has suggested that having a major is associated with persistence, but more recently Nora (2004) studying Latino students in three different educational contexts, including one HSI, concluded that a general academic fit was more important than college major.

Financial aid

Family finances and financial aid also predict degree completion, particularly grant aid for Latino and Black students (Cabrera et al., 1992; Perna, 2006; St. John & Musoba, 2010). Early research suggests that Latino students are loan averse and are likely to hold off-campus jobs and attend part-time in what some have called working-class enrollment patterns (Kaltenbaugh, St. John, & Starkey, 1999). While their study considered only transfer students at a HSI, Tuttle and Musoba (2013) found both grants and loans were associated with persistence.

Campus environment

The campus climates of HSIs have been associated with higher student persistence. Hagedorn, Chi, Cepeda, and McLain (2007), in a study of community colleges, concluded that a critical mass of Latino students increased the performance of Latino students. Maestas et al. (2007) demonstrated that traditional residential experiences such as living on campus, participation in sororities/fraternities, or holding a campus position positively affected students’ sense of belonging at a HSI. Nora (2004) found Latino students were more likely to reenroll if they felt personally accepted.

Family Encouragement

Multiple studies emphasize the importance of Latino families to their students’ persistence. Arana et al. (2011) reported Latino students at a HSI found power in their family educational advancement narratives and their first-generational status. The researchers concluded family support distinguished among Latinos in who persisted with more supported students being more likely to persist. Yet, other studies identified family responsibilities as burdensome and hindered Latino students’ success (Dayton et al., 2004; Medina & Posadas, 2012). While the research demonstrates the importance of the family role, our study did not have measures for family beyond measures of family income.

Clearly, more research is needed in HSI contexts that longitudinally examine the factors influencing student persistence. For this study, we are particularly examining student academic variables. Because first-year grades are important in prior research (Nora & Cabrera, 1996) and Santiago et al. (2004) suggests identifying gateway courses as a next step for HSI student persistence research, we decided to examine specific courses to determine if English composition and/or first-year math could be even more specific and early signals of students at risk of not graduating.

Method

This ex post facto study of retention at an HSI modeled bachelor’s degree attainment and followed a workable models approach, introduced by St. John (1992). He recommended using institutional records to control for student characteristics known to be associated with persistence while considering new variables of interest. Survival analysis has been used successfully in prior persistence and retention research (Chen & DesJardins, 2010; Ishitani, 2003) and is utilized here. Survival analysis as a statistical technique has been used most commonly in medicine to measure predictors of life expectancy after a medical event such as a heart attack. Here survival analysis is used to measure predictors of whether or not students graduate or drop out after beginning postsecondary education. We present a total of six analyses two each for the Latino, White, and Black student samples. We chose to exclude Asian and other students because the sample sizes were too small.

We were interested in the unique contribution of the key independent variables of first-year English and math coursework and first-year grades in predicting persistence within ethnic groups and not necessarily between ethnic groups. When considering all ethnic groups in one analysis, it is difficult to make claims about what is important for attainment for an ethnic subgroup without using too many interaction terms for our data set. Therefore, we conducted survival analyses for each ethnic group independently. However, this only allows limited comparison between ethnicities. It is inappropriate to interpret the size of the coefficients between models because they were separate analyses and widely varying sample sizes.

Data

Using institutional data, we sampled all first time in college students between the period of 2005-2010 who had applied for financial aid, whether or not they received it. This is not representative of the entire student body as there is a segment of foreign and U.S. students who cannot or choose not to apply for aid. We ran preliminary analyses including students who did not apply for aid and concluded those who applied for aid were not significantly different from the non-applicants in other key characteristics.

Our data sets included 3,304 Latino students, 522 White students, and 771 Black students from a predominantly commuter Hispanic-serving research university in a large urban area in the southeast. Most students come from within the local region, and the student body is only slightly older than traditional age.

Independent Variables

We considered a number of variables for inclusion based on the categories in the literature review, and our key variables were the college experience variables of first English and math courses taken and first-year grades. We also controlled for other postsecondary experience measures such as enrollment intensity and financial aid and precollege variables such as high school GPA and SAT scores and demographic characteristics. Considering the large number of initially selected variables, we took the reductionist econometric approach, reducing the number of variables based on non-significance in early regression models and retaining only those that were significant into the final analyses.

First English and first math courses were key variables as potential gateway courses. Math courses were either college algebra or pre-calculus, whichever the student took first and English composition for English. Both math and English milestones were measured through the final grade in the course, taking on values between 0 and 4. Student GPAs were left as continuous variables.

Other academic experience variables included first-term university GPA, part-time/full-time status, and total aid received. GPAs were left as continuous variables and enrollment intensity was coded as a dichotomous variable with 12 or more credits coded as full-time. Total financial aid received was in dollar amounts and left as a continuous variable. College major type and cumulative GPA were also considered but did not reach the level of significance to be included in the final regression models.

Precollege measures included high school GPA and SAT/ACT scores. In addition, we controlled for some demographic variables such as gender, citizenship, family income, dependence status for financial aid purposes, and age. We also considered parent education (first generation status) and high school type, but they did not reach the threshold of significance to be included in the final model. ACT scores were converted to SAT scores and a composite of the verbal and math SAT scores was used as a continuous variable. Gender, citizenship, and financial dependency status were coded as dummy variables with the largest groups as the reference group in the regressions; hence, female, U.S. citizens, and dependent on parent financial support were the omitted groups in the regressions.

Data Analysis

Realizing that our data set is both left and right censored, we approached our empirical specification with great care. The data were right censored in that some students were still enrolled at the end of the study, and the data were left censored in that we did not take into account at what point the student graduated, simply that they graduated. Censoring, together with possible violations of the spherical errors assumption (Cleves, Gould, Gutierrez, & Marchenko, 2008), guided our choice of survival analysis. Survival models are better equipped to handle censoring, and more importantly, they are better equipped to handle non-spherical errors and various hazard specifications than the alternative logit or probit models. Details about our model specifications and choices are available in the appendix.

Limitations

We did not have family income information for students who did not apply for financial aid and the expectation and receipt of aid can positively influence student behavior (Perna, 2006). However, as noted earlier, an analysis using financial aid application as a dummy variable covariate did not change the significance or direction of the impact of the other variables. We also did not control for institutional mentoring and off campus work or family responsibilities. Because we were using institutional data, these variables were unavailable. Considering prior research, these would have been useful measures.

Results

We present the results of the proportional hazard (PH) model (graduating) in detail, but only discuss the accelerated failure time (AFT) model (staying enrolled) when it adds new information or is inconsistent with the PH model. In the PH models, the coefficients indicate how covariates predict the hazard rate of graduating (where the baseline hazard coefficient is 1; Table 1), whereas in the AFT model they indicate their prediction on the survival time of not dropping out (Table 2). The two survival models (PH and AFT) have reverse coefficients and different interpretations. The term hazard can be confusing because in this situation, the hazard is a positive thing, earning a bachelor’s degree. The absolute differences in independent variables imply proportional differences at each time t. In addition, we report the distribution shape parameters ln p and p. We estimate the p parameter to be equal to 5.00 for Latino students, 7.91 for White students, and 7.02 for Black students—meaning the hazard rate is increasing with time for all groups, which is to be expected since the chance of graduating would increase as students take more classes. This essentially means that Latino students are (10/5)^5.00−1 (16 times) more likely to graduate after 10 semesters when compared with 5 semesters, whereas White students are (10/5)^7.91−1 (120 times) more likely to graduate after 10 than after 5. The differences between ethnic groups confirmed our need to analyze them separately.

Table 1.

Regression Results for Hazard of Graduating Using Weibull Parameterization.

	Results by ethnic group
Variable	Hispanic	White	Black
SAT score	0.9989294 (0.0007395)	1.000409 (0.001636)	1.00093 (0.0015211)
Gender	0.7406938* (0.099966)	0.5650834* (0.1860827)	0.7918551 (0.2320474)
Citizenship	1.121494 (0.1523347)	2.290498** (0.7901613)	0.5624412 (0.2483016)
Family income	1.000003** (1.22e-06)	0.9999997 (3.30e-06)	1.000005*** (1.95e-06)
Total aid received	1.000092*** (0.0000139)	1.000008 (0.0000375)	0.9999457* (0.0000308)
Dependency	0.2691477*** (0.0902983)	0.0792132*** (0.0599077)	0.4004398 (0.2639953)
Age	1.893022*** (0.0665743)	2. 224525*** (0.2198063)	2.274525*** (0.2128453)
English composition	1.248725** (0.1335146)	0.9416594 (0.2660138)	1.871896** (0.5028123)
College math	1.158129** (0.0673413)	0.9401193 (0.1406022)	1.647853*** (0.2188728)
Term 1 GPA	1.081763 (0.1356856)	1.774678* (0.589321)	1.538175* (0.3818366)
PT/FT indicator	0.8418368 (0.1131716)	0.8012126 (0.2594727)	1.592161* (0.4373103)
High school GPA	1.987953*** (0.3280671)	0.234274 (0.4695295)	0.9472254 (0.3117154)

/ln_p	1.610277*** (0.0859974)	2.068106*** (0.1565995)	1.948374*** (0.1627916)
P	5.004195*** (0.4303476)	7.909828*** (1.238675)	7.017269*** (1.142353)
N; LR (χ² df = 12)	3,304; 494.99	522; 97.18	771; 146.96

Note. Separate analyses for each ethnicity are presented in each column. Values greater than 1 are positively associated with graduating and values below 1.0 are negative. GPA = grade point average; PT/FT = part-time/full-time; LR = likelihood ratio.

α = .1. **α = .05. ***α = .01.

Table 2.

Regression Results Using Lognormal Parameterization Modeling Survival (Rather Than Graduating).

	Results by ethnic group
Variable	Hispanic	White	Black
SAT score	−0.000187 (0.0002043)	−0.0000456 (0.0002778)	−0.000321 (0.000246)
Gender	0.1260035*** (0.0392896)	0.0986397* (0.0555808)	0.0655746 (0.0501302)
Citizenship	0.0305743 (0.0387931)	−0.14513** (0.0661715)	0.0645337 (0.0678612)
Family income	−7.22e-07* (4.20e-07)	7.97e-08 (6.03e-07)	−7.77e-07* (4.02e-07)
Total aid received	−0.0000192*** (4.90e-06)	−3.05e-06 (6.47e-06)	5.83e-06 (5.17e-06)
Dependency	0.1243069 (0.0766826)	0.2831073** (0.1174953)	0.2199666*** (0.1296121)
Age	−0.2644234*** (0.0328707)	−0.1359546*** (0.029094)	−0.1498423*** (0.030614)
English composition	−0.0316061 (0.0290799)	−0.0238333 (0.0462038)	−0.0751228 (0.0471208)
College math	−0.0396242** (0.0159108)	0.017258 (0.0231888)	−0.0660963*** (0.0475801)
Term 1 GPA	−0.0737961** (0.0365407)	−.0361314 (0.0478881)	−0.0900695* (0.0768162)
PT/FT indicator	0.0130828 (0.0361069)	0.0385546 (0.0554776)	−0.0849285* (0.0491657)
High school GPA	−0.1514014*** (0.0463246)	−0.0624919 (0.0660548)	−0.0406551 (0.0554223)
Constant	9.860767*** (0.8694263)	6.202951*** (0.7804628)	7.258191*** (0.8616082)
/ln_sig	−1.014189*** (0.0803449)	−1.434577*** (0.1413367)	−1.486886*** (0.1347498)
Sigma	0.3626964*** (0.0291408)	0.2382162*** (0.0336687)	0.2260756*** (0.0304636)
N; LR (χ² df = 12)	3,304; 615.11	522; 93.86	771; 158.44

Note. Separate analyses for each ethnicity are presented in each column. Positive values are positively associated with surviving and values below 0 are negative. GPA = grade point average; PT/FT = part-time/full-time; LR = likelihood ratio.

α = .1. **α = .05. ***α = .01.

Academic Preparation

For Latino students, high school GPA positively predicted graduating, but not for Black or White students. Composite SAT score was not significant for any of the student ethnic groups in predicting persistence to graduation.

Postsecondary Academic Experience

First-year grades

Early coursework in math and English were important for all groups. For Latino students, success in their first math and first English composition courses significantly predicted graduation, but first-term GPA was not significant. For Black students, all first-year measures of achievement (Term 1 GPA and grades in first English and math courses) were significant and positively associated with persistence to graduation. Passing those math and English milestone classes increased the likelihood of graduation by 24% and 15%, respectively for Black students. For White students, the English and math courses were not significant; the only significant academic achievement predictor of graduating from this institution for White students was first-term GPA.

When looking at predicting drop out versus still being enrolled (AFT model), the first math course is significantly associated with continuing enrollment for Latino and Black students, but English composition is not significantly associated with being still enrolled for any of the groups. It appears that English composition distinguishes who graduates, but not who drops out. First-term GPA was associated with not dropping out for Latino and Black students but was not significant for White students.

Financial aid

Family income was important with higher incomes associated with better graduation rates for Latino and Black students. Financial aid was also associated with degree completion for Latino students. Latino students who received larger dollar amounts of financial aid in their first semester were more likely to graduate. This finding needed further exploration as all financial aid was bundled into one dollar amount, and many financial awards are merit based. Yet, we ran the entire set of regressions separating merit and need aid, but they did not yield any significant results for any ethnic groups (we do not show these insignificant results here). Specifically, only having both need and merit types of financial aid was predictive for timely graduation for Latino students. It may be that the dollar amounts of either alone were inadequate. Yet, selection for merit awards is non-random and endogenously correlated to the dependent variable of graduating. For Latino students, total aid received also significantly predicted being enrolled over having dropped out (AFT model).

For Black students, total financial aid received was negatively associated with graduating. Those students with the largest aid amounts also had the highest original need, and the aid may have been inadequate to support continuing enrollment. St. John and Starkey (1995) have suggested that a negative coefficient for financial aid can be interpreted to suggest that financial aid is inadequate. Contrary to the patterns for Black and Latino students, for White students, family income and financial aid were not significant, either positively or negatively. This suggests aid was adequate to make finances less important to their overall retention choices.

Enrollment intensity

Enrolling full- or part-time was not significant for Latino and White students. While counterintuitive, controlling for other factors, among Black students attending part-time was positively associated with graduating. This finding merits further research.

Discussion and Recommendations

Our results reiterate that standardized test scores do not predict who will graduate for students at a HSI, as did Arbona and Nora’s (2007) work. While SAT scores may predict early performance, and first-year GPA was in our model, these results also suggests that SAT scores may be less predictive of college performance in a HSI. This lack of consistency between research studies in the predictive power of SAT scores does not resolve the issues of the validity of the SAT (Shen et al., 2012), but further contributes to research that questions the appropriateness of using SAT scores in college admissions at HSIs.

For Latino and Black students, performance in their first English composition and first math courses were significant predictors of graduation. Latino and Black students may be using these courses as feedback for whether they believe they belong in higher education. Torres (2006) talks of Latino students looking for validation; and success, or lack of it, in key courses may be used as a form of validation by Latino and Black students. Math is a prerequisite course for most science, business, and engineering fields. Considering the low success rates in first math courses at universities in general and at this HSI, this is very problematic for Latino and Black students. We must offer academic support in these courses that is culturally appropriate. Some universities have seen success with more collaborative learning groups and group tutoring, which counters the unnecessarily competitive and individualistic nature of the math classroom (Robinson, 2007). Robinson showed that women were more likely to remain in math-related academic majors if they participated in group tutoring, and Robinson speculated this may be due to the collaborative learning process. More importantly, we must examine the way we approach instruction in these courses to improve student success (Becker, Ghenciu, Horak, & Schroeder, 2008). Culturally relevant pedagogy ties mathematic learning to students’ informal knowledge of mathematics, engages critical thinking, and builds on students’ cultural and experiential knowledge (Gutstein, Lipman, Hernandez, & de los Reyes, 1997; Ladson-Billings, 1997). The low success rates nationally in early math coursework suggests more research is needed at the postsecondary level on the teaching of mathematics, and application of research findings in the classroom is needed.

Somewhat surprising, first semester GPA was not a significant predictor for Latino students, as it was for other ethnic groups. This contradicts rigorous prior studies (Nora & Cabrera, 1996; Torres, 2006) that found first-year grades were very important to the ongoing enrollment of Latino students. It may be that the inclusion of the math and English composition courses represents some of the same construct. The greater specificity of the identified courses over an overall first semester GPA provides a clearer signal, and may allow institutions to intervene before the semester is over.

This article is only one piece of an ongoing process in understanding HSI student persistence. As our results suggest, it is important to consider subgroups of students within institutional contexts, as there may be differences in the way students respond to financial aid, institutional support, and signals about their academic performance. Students do look to the university for validation (Rendon, 1994; Torres, 2006), and our results suggest that at HSIs, students may see early course performance in math and English as forms of validation. Further research is needed on cost-effective ways institutions can provide students with validation and facilitate greater student success in early coursework.

Footnotes

Appendix: Additional Details About the Model Specification

We used a two-step verification procedure in choosing the best parametric survival model. First, we fit all of the alternative distributional specifications, as in equation (1). We then compared information criterions AIC (Akaike, 1973) and BIC (Schwarz, 1978) and chose the best model fit based on these criteria. Consequently, we also ran a semi-parametric Cox regression (Cox, 1972), which does not assume any distributional form for the errors, to compare those results to those of the chosen survival model. Getting similar magnitudes of the two model estimates suggested a proper choice of a parametric survival specification if a proportional assumption of the Cox model was justified. We modeled

h {(t)}_{j} = h_{0} (t) g (x_{j}),

where the individual hazard function $h {(t)}_{j}$ is a function of baseline hazard $h_{0} (t)$ and some set of covariates $g (x_{j})$ . Essentially, unspecified $g (x_{j})$ yields a semi-parametric Cox specification while the alternative specifications we considered were Exponential, Weibull, Gompertz, Lognormal, Loglogistic, and Generalized Gamma distributions in place of $g (x_{j})$ . It was immediately clear that Accelerated Failure Time (AFT) specifications performed strictly better in terms of the information criterions (AIC and BIC). This seemed consistent with the fact that the chance of graduation increases naturally as time evolves and students are closer to the goal. We found that, based on the AIC and BIC criteria, a lognormal distribution fit best. We adopted the ATF model with distribution $f (x) = \frac{1}{t σ \sqrt{2 π}} exp [\frac{- 1}{2 σ^{2}} {\ln (t) - μ}^{2}]$ and a survival function $S (t) = 1 - Φ {\frac{\ln (t) - μ}{σ}}$ . Hence, the hazard rate is defined by $h (t) = \frac{\frac{1}{t σ \sqrt{2 π}} e x p [\frac{- 1}{2 σ^{2}} {\ln (t) - μ}^{2}]}{S (t) = 1 - Φ {\frac{\ln (t) - μ}{σ}}}$ . We compared our estimates with thoseproduced using the popular proportional specification of a Cox model. We found predictions of parametric specification using a loglogistic distribution to be far off from the Cox model. This suggested one of two scenarios: First, the PH specification was wrong, and time should be treated in an accelerated fashion as in the AFT models. Alternately, the proportional assumption in fact was correct and the parameterization was wrong. Our ability to distinguish among the two possibilities was hindered by the fact that the proportional assumption can not be tested because of the number of time-varying covariates and late entries used in a Cox model estimation. A careful re-examination of the available parametric specification revealed that among the possible models (PH), the Weibull distribution produced the closest estimates, while scoring higher than the alternative (Exponential) PH model. As a comparison to the unassuming, a Cox model is a better check for the correct specification; therefore, we decided to adopt a Weibull distribution for the parametric specification. Hence, we adopted parameterization $λ_{j} = e x p (- x_{j} β)$ with ${e x p (- λ_{j} t_{j}^{p})}$ as a survival function in the PH model choice. We also chose to present the lognormal results for the AFT model. Figure A1 presents the smoothed hazard estimates for each ethnic group.

Authors’ Note

The views expressed in this publication are those of the authors and do not necessarily represent those of Lumina Foundation, its officers, or its employees.

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) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We wish to thank the Lumina Foundation for its generous support for this research.

Author Biographies

Glenda Droogsma Musoba is an associate professor of higher education at Texas A&M University. She earned her PhD from Indiana University and her research interests include educational policy, student success, and institutional practices.

Dmitriy Krichevskiy is an assistant professor of economics at Elizabethtown College (a selective liberal arts college). His primary research interest is entrepreneurship and self-employment with additional interest in education and student attainment. He earned his PhD in economics from Florida International University.

References

Akaike

(1973). Information theory and an extension of the maximum likelihood principle. In Petrov

B. N.

Csaki

(Eds.), Second international symposium on information theory (pp. 267-281). Budapest, Hungary: Akademiai Kiado.

Arana

Castañeda-Sound

Blanchard

Aguilar

T. E.

(2011). Indicators of persistence for Hispanic undergraduate achievement. Journal of Hispanic Higher Education, 10, 237-251.

Arbona

Nora

(2007). The influence of academic and environmental factors on Hispanic college degree attainment. Review of Higher Education, 30, 247-269.

Becker

Ghenciu

Horak

Schroeder

(2008). A college lesson study in calculus, preliminary report. International Journal of Mathematical Education in Science and Technology, 39, 491-503.

Bordes-Edgar

Arredondo

Kurpius

S. R.

Rund

(2011). A longitudinal analysis of Latina/o students’ academic persistence. Journal of Hispanic Higher Education, 10, 358-368.

Cabrera

A. F.

Castañeda

M. B.

Nora

Hengstler

D. S.

(1992). The convergence between two theories of college persistence. Journal of Higher Education, 63, 143-164.

Cabrera

A. F.

Nora

Castañeda

M. B.

(1993). College persistence: Structural equations modeling test of an integrated model of student retention. Journal of Higher Education, 64, 123-139.

Chen

DesJardins

S. L.

(2010). Investigating the impact of financial aid on student dropout risks. Journal of Higher Education, 81, 179-208.

Cleves

M. A.

Gould

W. W.

Gutierrez

R. G.

Marchenko

(2008). An introduction to survival analysis using Stata (2nd ed.). College Station, TX: Stata Press.

10.

Cox

D. R.

(1972). Regression models and life-tables. Journal of the Royal Statistical Society, 34, 187-220.

11.

Dayton

Gonzalez-Vasquez

Martinez

C. R.

Plum

(2004). Hispanic-serving institutions through the eyes of students and administrators. New Directions for Student Services, 2004(105), 29-40.

12.

Gutstein

Lipman

Hernandez

de los Reyes

(1997). Culturally relevant mathematics teaching in a Mexican American context. Journal for Research in Mathematics Education, 28, 709-737.

13.

Hagedorn

L. S.

Chi

Cepeda

R. M.

McLain

(2007). An investigation of critical mass: The role of Latino representation in the success of urban community college students. Research in Higher Education, 48, 73-91.

14.

Ishitani

T. T.

(2003). A longitudinal approach to assessing attrition behavior among first-generation students. Research in Higher Education, 44, 433-449.

15.

Kaltenbaugh

L. S.

St. John

E. P.

Starkey

J. B.

(1999). What difference does tuition make? An analysis of ethnic differences in persistence. Journal of Student Financial Aid, 29(2), 21-31.

16.

Ladson-Billings

(1997). It doesn’t add up: African American students’ mathematics achievement. Journal for Research in Mathematics Education, 28, 697-708.

17.

Maestas

Vaquera

G. S.

Muñoz Zehr

(2007). Factors impacting sense of belonging at a Hispanic-serving institution. Journal of Hispanic Higher Education, 6, 237-256.

18.

Mattern

K. D.

Patterson

B. F.

Shaw

E. J.

Kobrin

J. L.

Barbuti

S. M.

(2008). Differential validity and prediction of the SAT (Research Report No. 2008-4). New York, NY: The College Board.

19.

Medina

C. A.

Posadas

C. E.

(2012). Hispanic student experiences at a Hispanic-serving institution: Strong voices, key message. Journal of Latinos and Education, 11, 182-188.

20.

Nora

(2004). The role of habitus and cultural capital in choosing a college, transitioning from high school to higher education, and persisting in college among minority and non-minority students. Journal of Hispanic Higher Education, 3, 180-208.

21.

Nora

Cabrera

(1996). The role of perceptions of prejudice and discrimination on the adjustment of minority students to college. Journal of Higher Education, 67, 119-148.

22.

Otero

Rivas

Rivera

(2007). Predicting persistence of Hispanic students in their first year of college. Journal of Hispanic Higher Education, 6, 163-173.

23.

Perna

L. W.

(2006). Understanding the relationship between information about college costs and financial aid and students’ college-related behaviors. American Behavioral Scientist, 49, 1620-1635.

24.

Perna

L. W.

Thomas

S. L.

(2008). Theoretical perspectives on student success. ASHE Higher Education Report, 34(1). Retrieved from http://www.wiley.com/WileyCDA/WileyTitle/productCd-0470410787.html

25.

Rendon

(1994). Validating culturally diverse students: Toward a new model of learning and student development. Innovative Higher Education, 19, 33-51.

26.

Robinson

L. J.

(2007). Females, functions and finite: Women’s use of group mathematics tutoring and its impact on major choice (Unpublished doctoral dissertation). Indiana University, Bloomington.

27.

Sackett

P. R.

Kuncel

N. R.

Arneson

J. J.

Cooper

S. R.

Waters

S. D.

(2009). Socioeconomic status and the relationship between the SAT and Freshman GPA (College Board Research Report No. 2009-1). New York, NY: The College Board.

28.

Santiago

D. A.

Andrade

S. J.

Brown

S. E.

(2004). Latino student success at Hispanic-Serving Institutions. Available from www.edexcelencia.org

29.

Schwarz

(1978). Estimating the dimension of a model. Annals of Statistics, 6, 461-464.

30.

Shen

Sackett

P. R.

Kuncel

N. R.

Beatty

A. S.

Rigdon

J. L.

Kiger

T. B.

(2012). All validities are not created equal: Determinants of variation in SAT validity across schools. Applied Measurement in Education, 25, 197-219.

31.

St. John

E. P.

(1992). Workable models for institutional research on the impact of student financial aid. Journal of Student Financial Aid, 22(3), 13-26.

32.

St. John

E. P.

Musoba

G. D.

(2010). Pathways to academic success in higher education: Expanding opportunity for underrepresented students. New York, NY: Routledge.

33.

St. John

E. P.

Starkey

J. B.

(1995). The influence of prices on the persistence of adult undergraduates. Journal of Student Financial Aid, 25(2), 7-17.

34.

Tinto

(1993). Leaving college: Rethinking the causes and cures of student attrition (2nd ed.). Chicago, IL: University of Chicago Press.

35.

Torres

(2006). A mixed method study testing data-model fit of a retention model for Latino/a students at urban universities. Journal of College Student Development, 47, 299-318.

36.

Tuttle

L. V.

Musoba

G. D.

(2013). Transfer student persistence at a Hispanic serving university. Journal of Latinos and Education, 12, 38-58.