Menu Items

Overview: A quick overview of the menus.	The File Menu: Loading data,loading models, saving models.	The Distributions Menu: The distributions used in glmlab.
The Links Menu: The link functions used in glmlab.	The Scale Parameter Menu: The settings for the scale parameter.	The Residuals Type Menu: The types of residuals available in glmlab.
The Options Menu: The various options that can be set.	The Plots Menu: The residual plots that glmlab produces.	The Help Menu: The help available within glmlab.

Overview

Here are some pictures giving a quick look at the screen and the menus in glmlab. You may need to maximize the window to see everything.

The File Menu

The File menu contains these sub-menus:

Load Data File: An interactive window for loading data files;
Save glmlab Model: Saves the glmlab settings and the MATLAB workspace;
Load glmlab Model: Restores the glmlab settings and the MATLAB workspace from a previously saved model;
Close glmlab Window but Remember Settings: Closes the glmlab window, but remembers all the settings;
Quit glmlab and Lose Settings: Quits glmlab. I thought that was fairly obvious...
Exit MATLAB: See if you can guess this one...

The Distribution Menu

The Links Menu

The Link menu contains a list of the supported link functions: identity, logarithm, reciprocal, square root, to any power, logit (binomial only), probit (binomial only) and complementary log-log (binomial only). After these are listed any link functions defined by the user, if any.

Depending on the chosen distribution, a different default link is used. For the normal, the identity link is the default; for the Binomial, the logit is the default; for the Poisson, the log is the default; for the gamma, the reciprocal is the default. For user-defined distributions, the identity is the default.

If the binomial distribution is not selected, the logit, complementary log-log and probit links are dimmed, indicating that they cannot be selected.

The Scale Parameter Menu

The Scale Parameter menu has two options:
Mean Deviance: This estimates the scale parameter with the mean deviance This is the default for normal and gamma distributions, as well as user-defined distributions.
Fixed Value: This uses a fixed (positive) value for the scale parameter. This value can be chosen by the user, but the default is one. This is the default for the binomial and Poisson distributions. [Top of Page | Overview | File | Distributions | Links | Scale Parameter | Residuals | Options | Plots | Help]

The Residuals Type Menu

The Residuals menu determines what types of residuals are calculated. There are three options:

Pearson: Pearson residuals are the default;
Deviance: Deviance residuals;
Quantile: Quantile residuals. This option is not available for user-defined distributions. For a description of quantile residuals, see Dunn, P. K. and Smyth, G. K. (1996). ``Randomized Quantile Residuals'' in Journal of Computational and Graphical Statistics, 5, 1-10.)

The Options Menu

The Options menu has five sub-menus:

Declare New Model: After fitting models, select this options to reset glmlab for analysing a new set of data.
Restore Default Settings: Restores all the default settings (obviously...).
MATLAB Warning Message Status: Changes the status of the warning messages in MATLAB.
Change Fitting Parameters: Opens a dialogue box allowing the user to change fitting parameters (such as the maximum number of iterations, for example).
Include Constant Term in Model: If this option is checked, then a constant term is fitted to the model. If the item is unchecked, the model is fitted without the inclusion of a constant term.
Output Display: Allows the user to choose what information is displayed in the output.
Recycle Fitted Values: The algorithm uses the fitted values from the previous fit as starting values for the next fit.

The Plots Menu

The Plots menu is only enabled if residuals have been calculated. There are six items:

Residuals vs Response (y): Plots the residuals against the response variable.
Residuals vs Covariates (X): Plots the residuals against chosen covariates.
Normal Probability Plot of Residuals: Plots a normal probability plot of the residuals...obviously.
Residuals vs Fitted Values: Plots residuals against fitted values.
Residuals vs Transformed Fitted Values: Plots the residuals against a transformation of fitted values. The transformation transforms the fitted values to a constant information scale and depends on the error
distribution. See McCullagh and Nelder (1994), Generalized Linear Models (second edition) page 398 for more information.
Fitted Values vs Quantile Equivalents: Plots the fitted values against quantile equivalents of the residuals (not available for quantile residuals).

The Help Menu

The Help menu has eleven items:

On-line Manual: A link to the On-Line manual (that is, what you are reading).
Help with MAIN WINDOW Items: Gives a help screen for the input areas of the glmlab window.
Help with MENU Items: Help for the menus on the glmlab window.
Help with INTERACTIONS: A help screen for using interactions in glmlab (using the @ character).
Help with OUTPUT VARIABLES: Describing the output variables returned after fitting a model.
Help with RESIDUAL PLOTS: Describes briefly what the residual plots do.
Run glmlab Demo: Runs a fairly basic and short demo on the gui screen of how glmlab works.
Where to get glmlab: Places to obtain glmlab.
How to Contact the Author: How to get in touch with me if you would like.
Initial Splash Screen: Re-displays the initial welcoming splash screen that appears when glmlab starts.
Last Revision Date: Gives the date of the last revision to glmlab.

Last Revision: 09 June 1999

Author: Peter Dunn

E-mail: dunn@usq.edu.au