|
Activity Number:
|
283
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Education
|
| Abstract - #304045 |
|
Title:
|
Fitting and Interpreting GLMs (the Saturated Model Isn't Perfect and Other Cautionary Tales)
|
|
Author(s):
|
Robert Kushler*+
|
|
Companies:
|
Oakland University
|
|
Address:
|
24228 Edgemont Drive, Southfield, MI, 48033-6425,
|
|
Keywords:
|
Generalized linear models ; goodness of fit ; overdispersion ; negative binomial
|
|
Abstract:
|
Some issues that arise when using generalized linear models (GLMs) are discussed and illustrated. (1) Binomial and Poisson GLMs can be fitted using "grouped" or "ungrouped" versions of the data. Though the fit is identical, measures of practical significance can differ greatly. (2) The negative binomial model is useful for addressing the issue of overdispersion. An interesting computational issue occurs when fitting this model to a well-known data set.
|
