The items in this menu are repeated within the floating palette.
Define
Defines the Generalised Linear Model.
Errors
Normal, Poisson, Binomial, Gamma. Selecting each will cause the appropriate canonical link to be selected.
Convergence
The number of iterations of the fitting algorithm may be set. The default of 10 is normally sufficient but some data may require an increase. Failure to fit in 10 iterations usually means that something is wrong. For example your parameter estimates may be converging to + or - infinity. The convergence tolerance specifies the relative change in the deviance below which it is considered that the model has converged. Reducing this value results in more accurate estimates but may result in failure to converge. The default value represents a good compromise.
Residuals
Residuals may be determined using the following methods:
d(i) is the contribution to the deviance by observation i
Fit
Fits the model.
Use command-click (non-contiguous selection) or shift-click (contiguous selection) to select columns from list and then add these to the model using Add * (adds all hierachial models) or Add (adds the terms singly).
Estimates
Displays estimates and standard errors for the current model with p-values if these are selected in the preferences.
Residuals
Displays residuals for the current model. (Calculated using the method specified in the Define dialog)
Corr
Displays the correlation matrix for the currently fitted parameters.
Status
Displays current model settings and convergence parameters.