Analysis of Deviance for Generalized Linear Models
The Analysis of Deviance table for generalized linear
models, as illustrated by Figure 39.12,
includes the following:
- indicates the source of the variation.
Sources include Model for the fitted
regression and Error for the residual error.
C Total is the sum of the Model and Error components,
and it is the total variation after correcting for the mean.
When the model does not have an intercept term, the
uncorrected total variation (U Total) is printed.
- is the degrees of freedom associated
with each source of variation.
- is the deviance for each source of variation.
- is the deviance divided by its associated degrees of freedom.
When the scale parameter is a constant specified in
the method dialog, or when the response has a Poisson
or binomial distribution, the table also contains the following:
- Scaled Dev
- is the deviance divided by the dispersion parameter.
- Pr>Scaled Dev
- is the probability of obtaining a greater scaled deviance
statistic than that observed if the null hypothesis is true.
Under the null hypothesis, all parameters are 0
except for the intercept, and the scaled deviance
has an approximate distribution.
Copyright © 1999 by SAS Institute Inc., Cary, NC, USA. All rights reserved.