This five-day data modeling bootcamp consists of two parts.

Part 1: Machine Learning for Behavioral Modeling (May/June 2021, exact dates to be determined)

The first three-day training workshop covers the theory and application of machine learning (data mining, statistical learning) and exploratory approaches to data analysis. In contrast to traditional hypothesis-driven approaches to analysis, machine learning enables investigators to assess the predictive value of various combinations of variables in a data set. Machine learning has emerged in recent years as a major area of statistical research and practice and is increasingly employed by psychologists and other behavioral scientists. Machine learning techniques are particularly useful for the analysis of very large data sets, as can arise in clinical, survey, psychometric and genomic research. These techniques are often a natural follow-up to standard multiple variable and multivariate analyses where investigators have either: (1) obtained significant results and seek to know whether there are other important patterns in the data; (2) obtained no notable results and wonder whether there are any important patterns to be found; or (3) developed questions that are too general or imprecisely formulated to be addressed through hypothesis testing. The goals of this workshop are to provide attendees an understanding of various machine learning approaches, how to assess the utility of each approach, how to evaluate the predictive power of each approach, and how to tailor models to obtain replicable results. The workshop will include both theoretical and practical hands-on exercises using R, which is freely available at http://cran.us.r-project.org. Topics covered include an introduction to R, cross-validation using k-fold cross-validation, linear and nonlinear regression models, splines and smoothing splines, multiple regression, basic variable selection approaches in multiple regression (e.g., best subset regression, forward selection), advanced variable selection approaches in multiple regression (e.g., regularized regression, multivariate adaptive regression splines), logistic regression and decision theory, classification and regression trees, and ensemble methods (bagging, random forests, and boosting).

Part 2: Structural Equation Modeling (May/June 2021, exact dates to be determined)

The second part of the bootcamp constitutes a two-day workshop on Structural Equation Modeling. Structural Equation Modeling (SEM) techniques are used extensively in the social sciences to model behaviors. The strength of SEM lies in its ability to model a multitude of dependent variables simultaneously while explicitly considering error covariances across the equations that comprise the simultaneous equations model system. More importantly, SEM approaches are powerful methodological tools to consider attitudinal variables and constructs in behavioral modeling through the incorporation of latent variables that represent different attitudinal factors. By simultaneously embedding a factor analysis of attitudinal variables with a structural model of latent constructs, SEM is able to provide deep insights into behavioral relationships that connect attitudes and behavioral outcomes. This two workshop will cover the fundamentals of SEM including path analysis, exploratory and confirmatory factor analysis, latent variable path analysis, and multiple group analysis. The workshop will include both theoretical and practical hands-on exercises using R. Additional illustrations will be demonstrated using Mplus.

All registrants (of either workshop or the entire bootcamp) will receive a free copy of the following reference book.