Next: guipatterned() Up: Graphical Interface Help File Previous: guilistdlg()   Contents

guintrctplt()

Usage:
guintrctplt() does interaction plots via dialogs



Keywords:
guintrctplt brings up a dialog box to collect information that
will be used to construct an interaction plot command.  The
dialog is a "tabbed" dialog, meaning that the information is
collected on multiple panels that the user accesses via tabs.

Interaction plots graphically display a vector/matrix/array of
real numbers.  The coordinates of the first dimension are
indicated by the horizontal plotting position.  The levels of
any other dimensions are indicated by the plotting symbol; for
example, 2.4 indicates level 2 of the second factor and level 4
of the third factor.  Points with the same plotting symbol are
joined by lines.

The "Basic" tab determines the vector/matrix/array to be plotted.
First, you may directly select a matrix or array of plotting
positions by selecting a matrix of array in the variable selection
control at the bottom of the tab.  In this form, you only select
a single matrix or array.  Second, if there is an active model,
you may choose to plot least squares means from the active model.
To do this, you check the "use LS means" box in the "Control"
subdialog and select the desired factors from the model in the
variable selection control.  (Note: this form makes use of glmtable()
internally and thus will not work for balanced designs.  Use unbal:T
in the anova() command to enable the use of LS means in interaction
plots for balanced data.)  Finally, you may indicate the vector/
matrix/array to be plotted as the tabular means of a response 
variable split according to one or more factor variables.  In this
form, you first select the response variable, and then select one
or more splitting variables.

You may optionally choose to plot error bars around each mean, 
simply by checking the show error bars button in the "Control"
subdialog.  By default, the bars are plus or minus two SE, but you
may adjust that in the Control subdialog.  For LS means, standard
errors are taken directly from the model.  For the matrix/array
form, you must enter the name of a matrix of standard errors in
the "Options" subdialog.  The tabular data form will compute SEs
for the within-cell variances.  Note: an SE of 0 will be used
for cells with a single observation.  Optionally, checking the
"Pooled estimate of error" box in the Options subdialog will pool
variance information from all cells into a single common estimate
of variance, which will then be used to compute cell standard errors.
Finally, you may simply specify a common error variance directly
in the Options subdialog.

The "Appearance" tab collects information that affects the overall
appearance of the plot.  First, you can choose that the plot appear
in a new window (default), or you can choose the number of the
graph window where you would like it to appear. A window number of
0 indicates the most recently used graph window. Next, you can set
the width and height of the plot.  On the screen, these are in
units of pixels.  When printing using PostScript, these are in
units of points (approximately 1/72 of an inch).  

The second major set of choices are for labels.  You can add a title
and/or labels for the vertical and horizontal axes.

Finally, you can set where the border box and axis ticks will be
drawn.  By default, ticks and borders are drawn on all four sides.

The "Axes" tab allows you to control the appearance of the axes.
First, you can choose to have a logarithmic scale by clicking the
check box.  Next, you may specify your own minimum and maximum values
in each direction. Third, you may decide whether the x=0 or y=0 lines
are drawn on the plot.  Finally, you may set the appearance of the
ticks and labels.  Tick locations should be either a variable name
or an expression that evaluates to a vector of real values.  If
this is NULL, no ticks will be drawn.  Tick labels should be character
vectors with the same number of elements as the tick locations.
Finally, tick lengths should be real scalars >= -1.  Values less than
0 are outside the frame; values greater than zero are inside the
frame.  Values greater than 2 draw a grid all the way across the
plot.  The default value is -.5.

If you know the MacAnova commands, you may type in your options
directly on the "Direct Options" tab.


Gary Oehlert 2005-08-12