the functionality of four mixed-effects programs, MIXREG,
MIXOR, MIXNO, and MIXPREG, developed by Donald Hedeker
and Robert Gibbons into a single application to provide
estimates for mixed-effects regression models.
Mixed-effects models are
also known as multilevel, hierarchical, or random-effects
models. These models can be used for the analysis of longitudinal
data, where each individual may be measured at a different
number of occasions. They can also be used for clustered data,
such as for patients within clinics.
SuperMix has been developed
by SSI under an SBIR Phase II contract N44MH32056, and has been released in September 2008. SuperMix will fit models
with continuous, count, ordinal, nominal, and survival outcome
variables with nested data, allowing for up to three levels
of nesting. Key features are listed below.
- Easy to use graphical user interface: import data into the SuperMix spreadsheet, then build new models using menus and dialog boxes.
- Mixed-effects models for continuous outcome variables with auto-correlated residuals.
- Mixed-effects models for ordinal regression analysis, including non-proportional odds models and scaling effects.
- Mixed-effects models for Poisson regression analysis.
- Mixed-effects models for nominal logistic regression analysis.
- Mixed-effects models for grouped-time survival analysis.
- Two- and three-level models allowing for nested designs.
- Presentation quality graphics.