Logistic Regression Summary Download PDF
The following table lists the variables provided by the Summary port.
| Variable (Caption) | Description |
|---|---|
| Total samples | Total number of samples fed into the model |
| Total selected samples | Number of samples used in the model |
| Samples in the training set | Number of samples used to train the model |
| The root-mean-square error of the training set | The root-mean-square error indicates the model's accuracy on the training set. |
| Percentage of classification errors of the training set | Percentage of incorrectly assigned objects (observations or events) in a predefined class |
| The average cross-entropy of the training set | Cross-entropy provides a quantitative measure of the difference between two probability distributions. |
| Model training cutoff | Calculated value of the regression equation (ranging from 0 to 1) |
| -2 log-likelihood | A transformation of the likelihood function that indicates the prior probability of the regression model parameter values for a given value of the independent variable x. |
| Coefficient of determination | McFadden's coefficient of determination uses the log-likelihood to assess how well the regression model fits the actual data (ranges from 0 to 1). |
| Adjusted coefficient of determination | The adjusted form of McFadden's coefficient of determination is useful for comparing models with different numbers of factors (ranges from 0 to 1). |
| Chi-square | A chi-square statistic for the likelihood-ratio test. |
| Degrees of freedom for model | Number of independent model variables |
| Akaike information criterion | Use the Akaike information criterion to compare models with different numbers of parameters when you need to select the best set of explanatory variables. |
| Akaike information criterion corrected | Use this modified Akaike criterion for small samples, when the ratio of the number of samples to the number of model parameters falls below 40. |
| Bayesian information criterion | Another model selection criterion, based on the likelihood function and closely related to the Akaike information criterion |
| Hannan–Quinn information criterion | The Hannan–Quinn information criterion also uses the likelihood function and serves as an alternative to the Akaike and Bayesian criteria. |
| Model P-value | A statistic to assess the model's significance. The smaller the p-value, the less likely the null hypothesis that there is no statistically significant relationship between the dependent and independent variables. The system calculates it from the likelihood-ratio statistic. |
Read on: Logistic Regression Coefficients