Workspace Node: Cox Proportional Hazards - Specifications - Options Tab

In the Cox Proportional Hazards node dialog box, under the Specifications heading, select the Options tab to access the following options.

Estimation

Use the options in this group box to specify which likelihood (Breslow, Efron, or Discrete) to use in the presence of tied survival times.

Coding of factors

Use the options in this group box to specify how Statistica codes categorical variables in the design matrix. See Coding of categorical factors for more information.

Set reference category

Click this button to display the Category Codes dialog box, where you specify the reference category of each categorical factor.

Model Building

Use the options in this group box to specify how effects are entered into the model.

All effects

Select this option button to enter all effects specified in the current design.

Forward

Select this option button to specify the forward selection method. Effects are entered into the model one at a time beginning with no effects and ending when either there are no additional effects to enter or the effect associated with the largest adjusted score statistic does not meet the significance criterion defined by the p to Enter value.

Backward

Select this option button to specify the backward selection method. Effects are removed from the model one at a time beginning with the full model and ending when either there are no additional effects to remove or the effect associated with the least significant Wald statistic is less than the p to Remove value.

Stepwise

Statistica performs a stepwise selection by starting with the best fitting 1 variable model, and then sequentially adds variables according to the best adjusted chi-square score statistic if the p-value of the chi-square score statistic is less than the user specified p to Enter value. Unlike the forward selection method, an effect that has been entered into the model can be removed if the p-value of the Wald statistic is greater than the user specified p to Remove value.

Best subsets

Statistica performs a best subset search using the Branch and Bound algorithm of Furnival and Wilson (1974). The best subset for each number of predictors is found; that is, the best 1 predictor model is found, 2 predictor model, and so forth. Note that this option is only available for models that contain only covariates and no categorical factors.

p to Enter

Specify the criterion for an effect to enter in the model during a forward step. This field is available only when Forward or Stepwise is selected.

p to Remove

Specify the criterion for an effect to be removed from the model during a backward step. This field is available only when Backward or Stepwise is selected.

Effects to force

The Effects to force value enables you to force selected effects into the model (i.e., they will be part of every model that is considered); specifically, when the Effects to force value k is greater than 0 (zero), the first k effects in the design will be forced into all models that are evaluated.

Subject

This option can be used in conjunction with the selection of the robust covariance matrix, but it is not necessary to do so. In the counting process style of input, it might be common to have multiple rows or cases in the spreadsheet that pertain to a single subject. In order to aggregate results over each individual subject, a subject variable in the spreadsheet that identifies each subject can be specified.

Robust variance estimator

Select this check box to use the robust sandwich estimate of the covariance matrix.

Options / C / W. See Common Options.

OK. Click this button to accept all the specifications made in the dialog box and to close it. The analysis results are placed in the Reporting Documents node after running (updating) the project.