MIXOR & MIXREG: A Windows Computer Program for Mixed-Effects Regression Analysis with Autocorrelated Errors and Mixed-Effects Ordinal Regression Analysis Toolkit
Donald Hedecker & Robert D. Gibbons
This Toolkit includes Windows computer programs for a regression analysis of longitudinal and clustered data that are dichotomous, ordinal, or normally-distributed with autocorrelated errors. The programs provided employ a mixed-effects regression model, which assumes some regression parameters other than residuals to be random. Hence, the program adjusts for the dependence among observations present in longitudinal designs due to subjects being repeatedly assessed over time; and in clustered designs due to subjects being nested within a larger context, for example hospitals or firms. The mixed-effects regression model can be applied to the unbalanced data which, in longitudinal context, may mean unequal numbers of observations per subject. Or it can be applied to even observations at different points in time and in clustered context, unequal numbers of subjects per cluster. Both average outcomes for a population and individual outcomes for each subject can be estimated with the mixed-effects regression model. In addition to the computer programs, this Toolkit provides summaries on the computational methods used in the programs, the programs' descriptions, examples illustrating the programs' uses, and relevant journal articles.