Contour plots for linear models with quantitative regressors
This dialog allows to specify pairs of numeric regressors for which the response surface is to be displayed. Each display can be a contour plot with or without image colours, a 3D-perspective plot or an image plot.
For looking at several plots simultaneously, numbers of rows and columns can be specified. On OK, all plots are generated. One should always choose the number of rows and columns such that all requested plots fit on one display, otherwise the early plots will be overwritten by the later plots.
The functions behind the functionality are contour,
persp and image.
The second tab on the dialogue allows to specify at which values variables not in the plot are to be fixed. The default is the average for numeric variables, the first level for factors. (
These are used in the mfrow function for creating the
row and column layout for arranging several plots.
Within an R session, modifications are retained for further
calls of the dialog.
Contour plots show contours of constant height,
3D perspective plots show the surface in mesh form; both come with or
without coloring. The image plots
show colours only, without contours. It is mainly
a matter of taste which plots to use.
Within an R session, modifications are retained for further
calls of the dialog.
The default choice plots pairs for all numeric factors.
This can be too many plots for one page in case of many factors.
For the top radio button, pairs can be conveniently formed from two groups
of variables:
1: each pair within group 1
2: each pair within both groups 1 and 2 (but not pairs between groups)
3: each pair that involves any factor from group 1
4. each pair between groups 1 and 2 (but no pair within any group)
Initially, all variables are in group 1. They can be moved between groups
by selecting them with the mouse and moving them with the arrow buttons.
In case of the bottom (manual selection) radio button choice,
move selected 2-factor pairs between
the available list (those are not selected) and the selected list by
selecting them with the mouse (multiple selections possible) and
moving them with the arrow buttons.
These choices have to be redone with each call to the dialog.
Each response surface varies the values of two numeric variables, keeping all other variables fixed. These fixed values are also called the slice positions. The default slice positions are the average for numeric variables and the first level for factors. These can be modified on the second tab of this dialog. Within an R session, modifications are retained for further calls of the dialog w.r.t the same model.
Ulrike Groemping
See also contour for the functions
that do the actual plotting