Visualizing Categorical Data: halfnorm
$Version: 1.1 (9 Nov 2000)
Michael Friendly
York University
Half normal plot for generalized linear models
The HALFNORM macro plots the ordered absolute values of residuals from a
generalized linear model against expected values of normal order
statistics. A simulated envelope, correponding to an approximate 95%
confidence interval, is added to the plot to aid assessment of whether the
distribution of residuals corresponds to a goodfitting model.
The HALFNORM macro is called with keyword parameters. The RESP=
and MODEL= parameters are required. The arguments may be listed within parentheses in
any order, separated by commas. For example:
%halfnorm(resp=count, class=sex response, model=sexyearresponse@2);
 DATA=

Specifies the name of the input data set to be analyzed. The default is the
last data set created.
 Y=

 RESP=

Specifies the name of the response variable to be analyzed
 TRIALS=

The name of a trials variable, for
DIST=BIN
, with the data in events/trials form.
 MODEL=

Specifies the model formula, the righthadside of the MODEL statement. You
can use the  and @ shorthands.
 CLASS=

Names of any class variables in the model
 DIST=

Error distribution. The default is
DIST=NORMAL.
 LINK=

Link function. The default is
LINK=IDENTITY.
 OFFSET=

The
name(s)
of any offset variables in the model.
 FREQ=

The name of a frequency variable, when the data are in grouped form.
 ID=

The name of a character variable used as an observation identifier in the
plot...
 OUT=

Specifies the name of the output data set. The output data set contains the
input variables, absolute residuals (_ARES_), halfnormal expected value
(_Z_), Default: _RES_.
 LABEL=

Specifies whether and how to label observations in the plot.
LABEL=ALL
means that all observations are labelled with the
ID= variable value; LABEL=NONE
means that no observations are labelled; LABEL=ABOVE
means that observations above the mean of the simulations are labelled; LABEL=TOP
n means that the highest n observations are labelled.
 SEED=

Specifies the seed for the random number generators.
SEED=0
(the default) uses the timeofday as the seeed, so a different set of
simulated observations is drawn each time the program is run.
 RES=

The type of residual to plot. Possible values are: STRESCHI (adjusted
Pearson residual), STRESDEV (adjusted deviance residual),
Example
%include vcd(halfnorm); * or include in an autocall library;
%halfnorm();
See also
meanplot
panels
scatmat
stat2dat