Your Analytics.h++ library includes functionality for:
Multiple linear regression
The multiple linear regression features of Analytics.h++ include both parameter calculation and parameter statistics. The parameter statistics cover standard error, t-statistic and p-values, and confidence intervals. The library also includes a linear regression analysis of variance (ANOVA) class that provides quantities including degrees of freedom, mean square error, overall F statistic and p-values, and coefficient of determination. In addition, Analytics.h++ includes an F test class for testing hypotheses about model parameters.
Logistic regression
The library's logistic regression features provide parameter calculation and statistics. Parameter statistics include standard error, Wald chi-square statistic, p-values, and confidence intervals. Analytics.h++ also includes a fit analysis class that provides degrees of freedom; log likelihood values; deviance; Pearson and G statistic, and the Hosmer and Lemeshow statistic.
Model selection
Analytics.h++ provides the ability to search through a given set of predictor variables and select those that best describe the observed data. Best is determined by a user-supplied function object that evaluates the model for a given set of predictors. Analytics.h++ supplies the more common of such function objects, but you can create and use your own. Algorithms for searching the various combinations of predictor variables include forward selection, backward selection, stepwise selection, and exhaustive search.
©Copyright 1999, Rogue Wave Software, Inc.
Contact Rogue Wave about documentation or support issues.