About glmlab

• multiple regression;
• log-linear models;
• logistic regression; and
• weighted regression.

It was originally written to take the place of GLIM in a subject at the Department of Mathematics and Computing , University of Southern Queensland , but has since developed into a useful teaching and research tool in its own right.  glmlab can do many of the required tasks in generalized linear models through its GUI interface.

glmlab originally ran on MATLAB version 4 (the old version is still available from the MathWorks Web site), but now runs only in version 5.
 

The Features of glmlab

• normal (Gaussian);
• gamma;
• inverse Gaussian;
• Poisson; and
• binomial.

glmlab includes the following link function:

• identity;
• logarithm;
• reciprocal;
• square root;
• any power;
• logistic (binomial only);
• probit (binomial only); and
• complementary log-log (binomial only).

glmlab can automatically provide the following residual plots:

• residuals vs response;
• residuals vs any (or all) of the covariates;
• normal probability plot of residuals;
• residuals vs fitted values;
• residuals vs transformed fitted values (transformation to the constant information scale); and
• fitted values vs quantile equivalents.

• the parameter estimates (BETA);
• the standard errors of the estimates (SERRORS):
• the fitted values (FITS);
• the residuals (RESIDS);
• the covariance matrix of the parameter estimates (COVB);
• the covariance matrix of the differences between parameter estimates (COVD);
• the deviance at each iteration of the fit (DEVLIST);
• the linear predictor (LINPRED);
• the covariate matrix (XMATRIX); and
• the names of the variables fitted (XVARNAMES).

Some pictures:

• Click here for a look at the main window;
• Click here for an overview of what is under each menu item;

Downloading glmlab

• MathWorks Home Page
• MathWorks User-Contributed Files (for the latest "official" version)
• Local Updates ("non-official" versions of current version not sent yet to MathWorks) version 2.5:
• Local  tar file (best for UNIX versions)
• Local zip file (best for Windows versions)

---

The On-Line Manual

The manual probably isn't great, but I'm sure it will be adequate enough to learn enough of glmlab to get you going.

• Click here to go to the On-Line Manual
• Click here to download the pdf version of the entire manual, but be warned; it is quite large (635999 bytes actually).

Giving Feedback

There are two options for giving feedback:

1. E-mail me at dunn@usq.edu.au .
2. Write to me at:

Peter Dunn
Department of Mathematics and Computing
University of Southern Queensland
Toowoomba , Queensland 4350 AUSTRALIA

Cost

There are two options:

1. Write to me at:

Peter Dunn
Department of Mathematics and Computing
University of Southern Queensland
Toowoomba , Queensland 4350 AUSTRALIA
2. E-mail me at dunn@usq.edu.au .

  Comments about glmlab

If you want to add your own comments (but see above ), feel free to contact the author .

Links to Other Useful Pages

Generalized Linear Models

• Generalised Linear Models Page (Links to all kinds of issues to do with glm's, including a good introduction, and lists of software);
• Generalized Linear Models: An Introduction .

 MATLAB

•   Matlab (A very good introduction to using MATLAB);
•   The MathWorks Inc. - Developers of MATLAB & Simulink (The Home Page of the producers of MATLAB).
•   SStreams (Many links to good MATLAB sites, as well as other technical programs (Maple, S-Plus, Mathematica)).

Statistics

• Guide to the Web for Statisticians (a huge number of pointers to all sorts of resources, logically organized);
• Statistics on the Web ;
• StatLib .

Frequently Asked Questions

• Click here for the FAQ.

glmlab Papers

• A conference paper on glmlab was given at the 1997 MATLAB Conference in Sydney. Click here to go to the paper.
• The Journal of Statistical Software also published a paper on glmlab.  You can see it by clicking here .