Q: How can I calculate Statistics if SAAM II has not converged?

A: You might want to do this to have an idea of the local a posteriori identifiability properties of the model you are investigating. Statistical information is not available if the fitting has reached the maximum allowed number of iterations or if one or more parameters have reached their upper and lower limits (either in the subproblem or have actually hit them).

To calculate the Covariance Matrix and obtain all the statistical information even if the program has not converged, set the maximum number of iterations in the Computational Settings Window to zero. The program will halt after a single fitting run, the parameter values will be unchanged and the complete statistical information (Confidence Limits, Covariance and Correlation Matrix and Objective Function) will be made available in the Statistics Window.

Note that, even when using this approach, if a parameter (or more than one) has hit the Upper or Lower Limit, then its Statistics will not be available, and the other Statistics in the window will be evaluated keeping that parameter fixed to its current value and assuming for it a Standard Deviation of zero. Usually, this happens because the Confidence Limits on that parameter are very wide and the parameter estimate is therefore completely unreliable. Should you want the Statistics on that parameter anyway, and therefore have access to the complete statistical information, then its Upper and Lower Limits need to be relaxed (i.e. made wider, so that the parameter does not hit the Limits in the subproblem). In almost the totality of cases, if the Lower Limit is made mildly negative and the number of iterations is set to zero, the complete covariance matrix will be calculated and made available. You should, however, keep in mind that this kind of statistical information is for qualitative purposes only, and full convergence should be reached before putting any faith in the fitting results.