Generalized Linear Models
Fit Data with Nonnormal Response Distributions
Generalized Linear Models provide a unified way to fit responses that do not fit the usual requirements of least-squares fits. In particular, frequency counts, which are characterized as having a Poisson distribution indexed by a model, are easily fit by a Generalized Linear Model.
The technique, pioneered by Nelder and Wedderburn (1972), involves a set of iteratively reweighted least-squares fits of a transformed response.
Additional features of JMP’s Generalized Linear Model personality include the following:
• likelihood ratio statistics for user-defined contrasts, that is, linear functions of the parameters, and p-values based on their asymptotic chi-square ...