MIM is a program for graphical modelling. It is described in the book Introduction to Graphical Modelling. It is available for the Windows platform only, and can handle up to 52 variables.

It supports graphical modelling with both discrete and continuous variables, using undirected graphs, directed acyclic graphs and chain graphs. The latest release is MIM 3.2.

It is built around a comprehensive family of models for mixed discrete and continuous data, the hierarchical interaction models. Included as special cases are

• loglinear models for discrete data
• graphical Gaussian models for continuous data
• graphical interaction models for discrete and continuous data
• models for multiple continuous response variables with discrete and continuous covariates (multiple regression, ANOVA, MANOVA and MANCOVA models and generalizations of these)

In addition, a closely related family, the CG-regression models include

• binary and polytomous logistic regression models for a discrete response with discrete and continuous covariates
• models for multiple discrete and continuous responses given discrete and continuous covariates.

Maximum likelihood estimation is performed using the MIPS and TM algorithms. To incorporate missing data and/or latent variables, the EM algorithm can be used.

General likelihood ratio tests between nested models are supported. In addition specialized tests of conditional independence using a variety of tests, including exact tests, are available.

Several automatic and semi-automatic methods for model selection are implemented. These include stepwise methods, the EH-procedure, and selection based on information criteria.

The independence graph of the current model may be displayed dynamically in a popup window. The display may be edited to clarify the structure of the graph.

Various utilities include

• spreadsheet data entry and browsing
• data manipulation and calculation
• residual analysis
• Box-Cox transformations
• randomization tests
• simple data graphics - scatterplots, histograms, boxplots
• exact tests for ordinal categorical variables
• MIM can function as a COM automation server. This enables other programs to access it programatically.

For further information, see MIM's online help.

---

File translated from T_EX by T_TH, version 3.81.
On 19 Nov 2008, 14:55.