Research

IMPROVING METHODS FOR DATA-INTENSIVE RESEARCH

My goal is to provide applied researchers and organizations with statistical methods to answer research questions using complex large-scale data. The data sets I have worked with have thousands of cases (e.g., the Early Childhood Longitudinal Study), hundreds of thousands of cases (e.g., student data from the Florida Department of Education),  or millions of cases (e.g. student uses of the Algebra Nation system).  I have contributed to methodological advancement in the following areas: 1) Strengthening causal inference about program effects with large scale observational data (e.g., Leite at al., 2012); 2) Growth models for longitudinal data (e.g., Leite & Stapleton, 2011); 3) Estimation of nonlinear effects (e.g., Leite & Zuo, 2011); 4) Response bias and measurement error (e.g., Leite & Cooper, 2010) . I address methodological challenges through applications of structural equation modeling, mixture modeling, multilevel modeling (hierarchical linear modeling), and propensity score methods.

In the Fall of 2009, I created the Causal Analysis of Observational Studies research group. We are currently working on methods to combine propensity score matching with structural equation modeling and multilevel modeling to reduce selection bias in estimates of the effects of non-randomly assigned treatments in longitudinal studies.

IMPROVING METHODS FOR SCALE AND SURVEY DEVELOPMENT AND VALIDATION

I have provided several contributions to the field of psychometrics: I developed a new method for selecting items for short forms of scales based on the Ant Colony Optimization algorithm and structural equation modeling (Leite, Huang & Marcoulides, 2008); I proposed and evaluated a factor mixture model to identify individuals responding in a socially desirable way (Leite & Cooper, 2010), which is a common source of bias in educational and psychological research; I identified dimensionality problems of commonly-used social desirability bias scales (Leite & Beretvas, 2005). I have also engaged in multidisciplinary collaborations to address the validation of specific instruments: For example, I collaborated with Dr. Marilla Svinicki from The University of Texas at Austin to compare different multitrait-multimethod confirmatory factor analysis models for the scores of the VARK learning styles inventory (Leite & Svinicki, 2010).

I have served as principal investigator, co-principal investigator, or program evaluator for several grants which sum to over 6 million dollars in research funding. Details about my publications and grants are displayed in my Curriculum Vitae.