Cox Proportional Hazards Regression - Advanced Tab

Select the Advanced tab of the Cox Proportional Hazards Regression dialog box to access the options described here:

Convergence

This group box contains the options associated with how Statistica determines convergence of the Newton-Raphson algorithm.

Element Name Element Type Description
Maximum Number of Iterations Enter the maximum number of iterations for the iterative estimation procedure.
Absolute likelihood Statistica terminates the algorithm when the absolute value of the log partial likelihoods of the previous two iterations is smaller than the user-specified convergence criterion.
Relative likelihood Statistica terminates the algorithm when the value of the log partial likelihoods of the previous two iterations is smaller than the user-specified convergence criterion. Specifically, the criterion is defined as:

where eps is set to 0.000001

Relative parameter For each parameter estimate, the relative value of the current and previous estimate:

is computed (if previous estimate was 0, the absolute difference is computed). The maximum relative value is computed. This value is then compared to the convergence criterion.

Relative gradient The relative gradient criterion is defined as:

where g is the gradient vector, H is the Hessian matrix, and l is the likelihood function evaluated at the current iteration.

Convergence criterion Use this box to change the convergence criterion value.
Specify start values Check box

Select this check box if you want to specify your own starting values for the iterative estimation procedure. Then, click the Open data file button to specify the Statistica spreadsheet that contains the start values.

Open data file Button Click this button to display a dialog box where you can browse to the Statistica spreadsheet that contains the starting values. The number of rows in the spreadsheet should be equal to the number of parameters in the full model. There should only be one variable in this spreadsheet. Also, the order in which the parameters are specified is determined by the order of effects in the model. If you are unsure of the order, then run the initial model without user-specified start values, and click the Design Terms button on the Quick tab of the Results dialog box to produce a spreadsheet that contains the information for each column in the design matrix.
Edit Start Values Button Click this button to display the Edit start values and specify to estimate/fix params dialog box, where you can edit the start value for each parameter with the additional option to set a parameter as constant. If the parameter is specified as constant, Statistica does not estimate the parameter via maximum likelihood, but instead leaves the parameter at its user specified value.