Iteration Results

Click OK (Run Model) in the Startup Panel to display the Iteration Results dialog box, which reports the progress of iteration. The information reported can be useful in diagnosing problems in convergence. When iteration is terminated, you are presented with two options that give you the opportunity to continue the analysis or terminate without presenting the Results dialog box. Note that the Iteration Results dialog box is also displayed during the Monte Carlo Analysis iteration process.

When iteration stops, the following commands are available:

Cancel

Click the Cancel button to terminate the current analysis and return to the Startup Panel.

OK

When you click the OK button, STATISTICA calculates results and displays the Results dialog box.

Iteration Results

The following information is displayed in the Iteration Results dialog box:

Itn.  #.

The number of the iteration in the iterative sequence.  

Discrepancy

The current value of the discrepancy function being minimized.

RCos

The current value of the Maximum Residual Cosine criterion.

Lambda

The value of the step multiplier used on that iteration. A value of 1. means that the first "full" step reduced the discrepancy function "sufficiently" to continue iteration. A value less than one means that the program had to use a line search along the chosen step direction in order to find a point where the discrepancy function was reduced. Very small values usually indicate that iteration is about to fail.

MAXCON

The maximum value of any constraint function. This value will only be non-zero during constrained estimation, used when the New option is chosen for Standardization, or Correlations are chosen as the Data to Analyze. If iteration progresses satisfactorily, this value should decrease systematically to a value very close to zero.  

NRP

Number of redundant parameters. If some parameters are redundant with others, the program may be able to detect the condition, and, if so, will set this value other than zero.

NRC

Number of redundant constraints. If some parameter constraints are redundant, the program may be able to detect the condition, and, if so, will set this value other than zero.

NAIC

Number of active inequality constraints, or "boundary conditions." During iteration, the program maintains certain inequality constraints on parameters to prevent some "impossible" values from arising. For example, negative variances are not allowed to occur. If the program detects a variance will be set to a negative value on the next iteration, it constrains the value at the boundary point (i.e., zero) and only minimizes the discrepancy function relative to the other parameters. If, for example, you have a "Heywood case" in a confirmatory factor model, you will have a non-zero value of this index at termination of iteration.

StepLen

The length of the current full iteration step. If an asterisk appears next to the value, it indicates that the maximum allowable step was taken. You can control the maximum step length allowed by setting a parameter in the Analysis Parameters dialog box.