After a model has been estimated, you may wish to test hypotheses about the values of coefficients with reference to each other. For example, you may wish to test whether two or more coefficients are equal to each other, or whether they sum to one.
Wizard supports four kinds of hypothesis tests for model coefficients. Each is associated with a null hypothesis, as described in the table below.
|Hypothesis test||Null hypothesis||Formula|
|Joint significance||Coefficients are zero||βi=0|
|Coefficient equality||Coefficients are equal||βi=βj|
|Sum-to-zero test||Coefficients sum to 0||∑βi=0|
|Sum-to-one test||Coefficients sum to 1||∑βi=1|
To test one of these hypotheses about two or more model coefficients:
Wizard performs all hypothesis tests using the Wald test. An F statistic is constructed for linear models, and a chi-squared statistic is constructed for non-linear models.