Purpose of Nonlinear Estimation

Select the Advanced tab of the Nonlinear Estimation Results dialog box to access the options described here.

Summary: Parameter estimates. Click the Summary: Parameter estimates button to create a spreadsheet with the parameter estimates for the current model. Summary statistics for the current analysis will be displayed in the spreadsheet header.

Note: If
the Asymptotic standard errors
check box is selected on the Model Estimation dialog box -
Advanced tab, the Summary:
Parameter estimates button is replaced by the Summary:
Parameters & standard errors button. Click this button to display
a spreadsheet with the parameter estimates for the current model as well
as the (asymptotic) standard errors of the parameter estimates, the respective t-values (parameters divided by standard
errors), and the associated p-values.
Note that standard errors are always computed when User-specified
regression, least-squares is selected from the Startup Panel.

Logistic regression. Several additional summary statistics are displayed in the spreadsheet when logistic regression is performed. Odds ratios are displayed for each parameter estimate for a unit change in the predictor variable and for a change equal to the observed range of values of the predictor variable. If the Asymptotic standard errors check box is selected via the Model Estimation dialog box - Advanced tab, then the spreadsheet will also show Wald's Chi-square statistics and associated p-values for testing the significance of the parameters. Upper and lower confidence limits for the parameter estimates and the odds ratios will also be displayed. These confidence limits are computed using the Alpha value specified in the Confidence intervals for parameter estimates box.

p-value for highlighting. Adjust the p-value by entering a new value in the p-value for highlighting box or by using the microscrolls. The default p-value for highlighting is .05. On the summary spreadsheet, STATISTICA will highlight all p-values that are equal to or less than the value specified in this field. For more details on p-value, see Elementary Concepts.

Scale MS-error to 1. Select the Scale MS-error to 1 check box to rescale the mean square error to 1, which is recommended for maximum likelihood estimates. The resulting standard deviations for the parameter estimates are then the usual information theory standard errors (Jennrich & Moore, 1975). When using maximum likelihood estimation for probit or logit models, this check box Is automatically selected.

Confidence intervals for parameter estimates. Adjust the confidence intervals for parameter estimates by entering a new value in the Confidence intervals for parameter estimates box or by using the microscrolls. The default value is 95%.

Covars & correlations of parameters. Click the Covars & correlations of parameters button to produce a spreadsheet with the (asymptotic) variance/covariance matrix for the parameter estimates and a spreadsheet with the correlations between parameter estimates. Note that the Covars & correlations of parameters button is only available if you select the Asymptotic standard errors check box on the Model Estimation dialog box - Advanced tab.

Difference from previous model. Click the Difference from previous model button to display a spreadsheet comparing the goodness-of-fit between the current and previous models. The Difference from previous model button is only available if logit or probit regression is selected and the current model is hierarchically related to the previous model that was estimated. "Hierarchically related" means that the current model is identical to the previous model with the exception of an addition or deletion of one or more independent variables.

Fitted 2D function & observed vals. Click the Fitted 2D function & observed vals button to create a two-dimensional graph of the observed values. This plot allows a two-dimensional visual examination (i.e., qualitative evaluation) of the fit of the data to the model. It is useful for identification of outliers, which can then be marked in the Data Editor using the Brushing Tool button.

Fitted 3D function & observed vals. Click the Fitted 3D function & observed vals button to produce a three-dimensional graph with the observed values. This plot allows a three-dimensional visual examination (i.e., qualitative evaluation) of the fit of the data to the model. It is useful for identification of outliers, which can then be marked in the Data Editor using the Brushing Tool button.