This module computes a basic version of the Generalized Linear Model, for a single explanatory variable. It requires two columns of data (independent and dependent variables).
GLM allows non-normal distributions, and also “transformation” of the model through a link function. Some particularly useful combinations of distribution and link function are:
- Normal distribution and the identity link: This is equivalent to ordinary least squares linear regression.
- Normal distribution and the reciprocal link: Fit to the function y=1/(ax+b).
- Normal or gamma distribution and the log link: Fit to the function y=exp(ax+b).
- Binomial (Bernoulli) distribution and the logit link: Logistic regression for a binary response variable (see figure above).
For mathematical details, see the Past manual.