Teacher Certification Tests: Logistic Regrusing Linear and Ession Models to Predict Success of Secondary Pre-Service Teachers
Poelzer, G. Herold, Zeng, Liang, Simonsson, Marie, College Student Journal
This research, conducted at a university in South Texas (87% Hispanic), investigated how well pre-service teachers' performance on the Secondary Professional Development Examination for Certification of Educators in Texas (ExCET) can be predicted. The linear regression equation explained 51% of the total variance in the ExCET scores while the logistic regression equation accurately predicted 80% of the overall pass/fail status. This study concludes that English language skills contribute enormously to the success of Hispanic pre-service teachers on the ExCET and suggests that school districts ensure that students develop these skills. The methodology used in this study is applicable to certification tests in general.
For decades, politicians, educational leaders, educators, and researchers have discussed and debated issues pertaining to educator standards and certification (Riley, 1999; Darling-Hammond, 1995; White, Burke, & Hodges, 1994; Chambers, Munday, Sienty, & Justice, 1999; Jaeger, 1988; Cornett, 1987; Stinnett, 1967). These activities have initiated changes in legislative mandates and accountability systems that, in turn, have brought about much educational reform nationwide (Cornett, 1987).
In 1987, for example, Texas replaced the Bachelor of Education Degree with a Bachelor of Interdisciplinary Studies Degree, a degree that requires an academic major along with 18 semester hours of education courses (including student teaching) and, for certification, a passing grade on each of a series of comprehensive examinations testing knowledge in both subject content and pedagogy (Cornett, 1987). Several recent changes in Texas occurred in specific grade level certificates, certification examinations, and accountability.
Currently, the grade levels for which teachers require specific certificates consist of early childhood (EC) through 4th, 4th through 8th, and 8th through 12th grade, and various all-level certificates, replacing an EC through 6th or 8th grade, 6th through 12th grade, and a smaller number of all-level certificates (National Evaluation Systems, 2004). In addition, the Texas Examinations of Educator Standards (TExES), introduced in 2002, phased out the Examination for Certification of Educators (ExCET) on August 31, 2005. Furthermore, the state of Texas has become the first state to establish a K-16 accountability system. What's more, Texas institutions of higher education with teacher preparation programs have been held accountable for the performance of pre-service teachers on state required certification exams since 1998 (San Miguel, Garza, & Gibbs, 2000).
The purpose of this study was twofold: (1) to identify variables that predict success on the Secondary Professional Development ExCET, and (2) to demonstrate a methodology and statistical models that are cogent tools for identifying variables that predict success on certification tests in general.
Researchers at two predominately White universities in Texas have identified variables that predict success on the Secondary Professional Development ExCET: Chambers, Munday, Sienty and Justice (1999) at Texas A & M-Commerce not only identified critical thinking as an important predictor variable but also found that the combination of reading ability, grade point average, age, gender, and the subscales of the Texas Academic Skills Program (TASP) math, reading, and writing scores can be used as predictors. White, Burke, and Hodges (1994) at Lamar University found SAT scores and grade point average to be good predictors. In more recent studies, researchers at a mainly Hispanic university in South Texas identified other variables that predict success: benchmark scores, TASP reading scores, and ACT composite scores (Poelzer, Zeng, & Simonsson, 2000; Simonsson, Poelzer, & Zeng, 2000; Zeng, Poelzer, & Simonsson, 2002). Poelzer et al. primarily used linear regression equations to identify predictor variables; however, this study goes further by employing a model that includes both linear and logistic regression equations. …