Research Interests

I have two main lines of research: the design of research studies and longitudinal data analysis (i.e., the analysis of change). My specific interests within the area of research design is primarily focused on issues of sample size planning. In particular, I focus on the interplay between sample size planning for statistical power (i.e., power analysis) and sample size planning for statistical accuracy (i.e., accuracy in parameter estimation, AIPE). Statistical accuracy in this sense refers to obtaining confidence intervals that are sufficiently narrow.  A  narrow confidence interval provides more information about the population parameter of interest than does a wide interval or a null hypothesis significance test, as the interval reveals whether or not some null value (generally zero) can be rejected and it defines the range of plausible values for the parameter (which I argue is ultimately what is of interest). Depending on the specific question of interest, planning a research study so that a specified level of statistical power is realized may not address the particular question of interest. Many times obtaining a narrow confidence interval will better address the research question, such as when the question relates to the value of the population parameter of interest (which is often the case). Thus, rather than planning sample size so that a false null hypotheses can be rejected, sample size can be planned so that the uncertainty of the population parameter(s) of interest can be minimized by obtaining a confidence intervals that are narrow (thus probabilistically revealing information about the size of the effect). Nevertheless, in some situations rejecting a false null hypothesis is exactly what is needed to answer the research questions. Realizing this, I also work on sample size planning from a power analytic perspective. Much of this research deals with effect sizes and their corresponding confidence intervals. MBESSis an R package that allows researchers to plan the appropriate sample size for power analysis and AIPE for a wide variety of effect sizes (both standardized and unstandardized). MBESS has been peer reviewed in Behavior Research Methods and Journal of Statistical Software, and methods from MBESS are included with several of my peer reviewed publications (see the references).  

My specific interests withing the context of longitudinal data analysis are on the use of nonlinear models (i.e., mixed models or hierarchical nonlinear models) of change and modeling change in situations where hypothesized but unknown heterogeneity exists. Methods for modeling and understanding change in the context of longitudinal designs are important because with repeated measures data, a rich set of hypotheses can be addressed. However, repeatedly measuring the same set of individuals leads to problems and complications not realized in cross sectional research. I try to find ways in which these problems can be overcome and to develop methods so that particular questions of interest that deal with change can be addressed in an optimal way. A wide variety of methods exist for modeling longitudinal data, but not all of the methods are optimal or even useful for certain research questions. A goal of my research is to develop, improve, and apply methods for studying longitudinal data so that important questions can be addressed in a statistically optimal way.

Actually, my interests span widely across the field of research methodology. Some of the other topics that I work on or have a real interest in are general latent variable modeling (e.g., structural equation modeling, growth mixture modeling), multilevel models (especially models nonlinear in their parameters useful in an analysis of change context), the theoretical and conceptual arguments of using effect sizes and confidence intervals for effect sizes, finite mixture modeling, statistical classification and statistical discrimination, and computational statistics (e.g., the bootstrap technique and the proper design and implementation of Monte Carlo simulation studies). I also dabble in measurement issues from time to time, but I devote nearly all of my time to statistical issues. The methods that I am interested in need not be conceptualized as being mutually exclusive. In fact, a major area of my research involves the combination of these techniques so that substantive questions can be addressed with methodologically sound techniques and procedures, and so that existing methods can be extended and generalized.

Collaboration

I've been a consultant on many research projects ranging from small scale narrowly focused studies to large scale government funded projects. Feel free to contact me if you think my research could be beneficial to your research. Depending on many factors, I may or may not be able to provide assistance. 

Get Involved

If your interested in working on methodological issues that arise in the behavioral, educational, or social sciences, feel free to contact me. This is especially true for undergraduate students considering a graduate program (or graduate students interested in changing programs). Highly motivated advanced undergraduate students at Indiana University who are interested in getting involved with research on statistical methods should also feel free to contact me.