GLM, GRM, and MANOVA Syntax - Keyword Estimate

Optional keyword; specify the coefficients that are to be used in the linear combination of parameter estimates for the custom hypothesis; multiple ESTIMATE specifications may appear in the same analysis. Note that tests of linear combinations of parameter estimates can also be requested from the GLM Results dialog, where a convenient and efficient user-interface is provided for specifying the coefficients.

Continuous predictors

The number of parameters specified in the Value list must match the number of parameters in the respective model. In the example shown above, if a multiple regression model was specified with the intercept (X0) and three continuous predictor variables (X1, X2, X3), then for each linear combination four (intercept plus three predictors) parameters must be specified (b0, b1, b2, b3); the statement shown above

0*b0+1*b1-1*b2+0*b3 = 0

0*b0+1*b1+0*b2-1*b3 = 0

Thus, these coefficients will test the hypothesis that the parameter estimates for X1, X2, and X3 are identical.

Categorical predictors

In models involving categorical predictors, the interpretation of the parameter estimates, and thus the coefficients in the ESTIMATE command depends on the parameterization of the model (overparameterized or sigma-restricted). If the overparameterized model is specified, the ESTIMATE command can be used to specify planned comparisons between selected cells in the design. Refer to Hypotheses about linear combinations of effects and Testing specific hypotheses for additional details. In either case, the number of coefficients following the ESTIMATE command must match the number (and order) of parameters in the model; remember that missing cells are dropped from the design matrix, and thus the corresponding parameters will also be dropped. In complex designs, it is easiest to use the Estimate option on the GLM Results dialog to construct custom hypotheses.

Applies to

GLM, GRM, and ANOVA/MANOVA