Overview (GLM command)

GLM (general linear model) is a general procedure for analysis of variance and covariance, as well as regression. GLM is the most versatile of the analysis-of-variance procedures and can be used for both univariate and multivariate designs. GLM allows you to:

  • Include interaction and nested effects in your design model. Multiple nesting is allowed; for example, A within B within C is specified as A(B(C)).
  • Include covariates in your design model. GLM also allows covariate-by-covariate and covariate-by-factor interactions, such as X by X (or X*X), X by A (or X*A), and X by A within B (or X*A(B)). Thus, polynomial regression or a test of the homogeneity of regressions can be performed.
  • Select appropriate sums-of-squares hypothesis tests for effects in balanced design models, unbalanced all-cells-filled design models, and some-cells-empty design models. The estimable functions that correspond to the hypothesis test for each effect in the model can also be displayed.
  • Display the general form of estimable functions.
  • Display expected mean squares, automatically detecting and using the appropriate error term for testing each effect in mixed-effects and random-effects models.
  • Select commonly used contrasts or specify custom contrasts to perform hypothesis tests.
  • Customize hypothesis testing, based on the null hypothesis LBM = K, where B is the parameter vector or matrix.
  • Display a variety of post hoc tests for multiple comparisons.
  • Display estimates of population marginal cell means for both between-subjects factors and within-subjects factors, adjusted for covariates.
  • Perform multivariate analysis of variance and covariance.
  • Estimate parameters by using the method of weighted least squares and a generalized inverse technique.
  • Graphically compare the levels in a model by displaying plots of estimated marginal cell means for each level of a factor, with separate lines for each level of another factor in the model.
  • Display a variety of estimates and measures that are useful for diagnostic checking. All of these estimates and measures can be saved in a data file for use by another procedure.
  • Perform repeated measures analysis of variance.
  • Display homogeneity tests for testing underlying assumptions in multivariate and univariate analyses.