5.3 Parameter Estimate Classes

Analytics.h++ provides container classes for storing regression model parameter estimates and their associated statistical quantities. These classes are:

RWLinearRegressionParam, for multiple linear regression parameter estimates

RWLogisticRegressionParam, for logistic regression parameter estimates

Both classes contain parameter estimates: t statistics (for linear regression), Wald chi-square statistics (for logistic regression), p-values, critical values, and confidence intervals.

Instances of these classes are constructed only by their corresponding regression classes. For example, to obtain the vector of parameter estimates for a logistic regression, you must construct an instance of RWLogisticRegression and invoke the parameterEstimates() method on the object, as demonstrated in the following example:

RWGenMat<double> predictorMatrix;       // Values of the predictor
                                        // variables.
RWMathVec<RWBoolean> observationVector; // Values of the dependent
                                        // variable.
.
.
.
RWLogisticRegression lr( predictorMatrix, observationVector );
RWTValVector<RWLogisticRegressionParam> params =
                                          lr.parameterEstimates();
.
.
.

There are also classes for testing hypotheses about linear regression model parameters.