Report
The variables provided by the Summary port are represented in the table:
| Variable (Caption) | Description |
|---|---|
| Total samples | Total number of the samples supplied to the model input |
| Total selected samples | Number of the samples used in the model |
| Samples in the training set | Number of the samples which are used for the model training |
| The root-mean-square error of the training set | The training error that reflects the model setting accuracy of the training set |
| Percentage of classification errors of the training set | Percentage of incorrect assignment of objects (observations, events) to one of the classes known in advance |
| Average cross-entropy of the training set | Quantitative evaluation of difference between two distributions of probabilities |
| Model train cutoff | Calculated value of regression equation (changes from 0 to 1) |
| -2 Logarithm of the likelihood function | The function that defines occurrence probability of values of the regression model parameters for the set value of independent variable x |
| Determination coefficient | Value of link between variables of regression model (changes from 0 to 1) |
| Adjusted determination coefficient | Value of link between variables of regression model (changes from 0 to 1) It differs from not adjusted determination coefficient by the fact that the adjusted determination coefficient can be decreased when entering the variables that do not have considerable impact on the dependent variable into the regression model. |
| Chi-square | Chi-squared test to test the hypothesis concerning the law of distribution of the random value under study |
| Number of freedom degrees of the model | Number of independently varied values of indicator |
| Akaike information criterion | The criterion is used for comparison of the models with different number of parameters when it is required to select the best set of explicative variables |
| Akaike information criterion corrected | The modified Akaike criterion used for the small samples when ratio of the number of the samples in it to the number of the model parameters is less than 40 |
| Bayesian information criterion | The criterion is based on use of likelihood function and it is closely connected with the Akaike information criterion |
| Hannan-Quinn information criterion | Alongside with the Akaike and Bayesian criteria, it is specified in the results of assessment of the models with discrete and limited dependent variables |
| Model P-value | Assessment of the model accuracy. It is probability of the fact that the test statistics value of the used criterion calculated for a sample will exceed the set P-value |