Applied Statistics for Forensic Psychologists
Full course description
The course consists of six units.
The first unit will focus on a review of multiple linear and logistic regression analysis, which will form the basis for most of the advanced techniques that will be covered in the remainder of this course. This general introduction into regression techniques is followed by a unit that focuses on ROC curve analysis. ROC curves are becoming increasingly more important to forensic psychologists, for instance, to help find optimal cut-off scores for instruments that should help decide on whether an institutionalized offender can be granted parole or leave, or for studying whether verbal veracity assessment tools can discriminate between truth or falsehood of incriminating or exculpatory statements. In studies on the accuracy of identification of crime suspects, a comparison of ROC curves can be used to decide which of several line-up procedures is superior in terms of maximizing correct and minimizing false identification rates. The third unit will cover meta-analysis.
The last three units are devoted to mixed (multilevel) regression for nested designs and longitudinal studies. This mixed regression starts with a unit on marginal models for repeated measures (for instance, a time series of observations on institutionalized offenders receiving specialized treatment). Especially in cases of missing data or within-subject covariates, such models are known to be more efficient than traditional techniques such as repeated measures ANOVA. In this first of three units on multilevel regression, students are shown the pros and cons of various models for the correlational structure of repeated measures, such as compound symmetry and AR1. The second unit covers the random intercept model for repeated measures as a method to include individual effects in marginal models for longitudinal data (growth curves) or single trial analyses of lab data. Students learn how this can be combined with e.g. ARMA modelling to distinguish between interpersonal and intrapersonal outcome variation. The random intercept model will also be applied to a cluster randomised trial - i.e. an RCT where organisations, like institutions treating justice-involved adolescents, are randomised. The third and last unit on mixed regression covers random slope models for longitudinal data (individual differences in change over time), single trial analysis (individual differences in stimulus effects) and multicentre trials (RCT within each of a number of organisations).
- students are able to identify and apply appropriate regression models for continuous and binary outcome variables, for data with independent as well as with correlated residuals;
- they will also be able to explain the use of ROC curves and will understand how to apply these in the context of classification on the basis of test scores;
- students will also understand how to apply basic multilevel analysis for both longitudinal and nested data;
- they will be able to explain the key concept of a covariance structure;
- lastly, students will be able to explain the basic concepts used in meta-analysis, and to read and interpret basic output from the Metafor-package in R.
Good understanding of descriptive and inferential statistics at the elementary and intermediate level, including multiple linear regression. Skilled in the use of SPSS for statistical data analyses.