SAS Helpful Hints
Mass Production of Proc GLM
%Anovas resides in the departments SAS
Macro Library and is enabled with the following SAS statements:
L ibname
biostat '\\BIO2\example\SASMacrosWindows';
Options Mstored SasMstore=biostat;
%Anovas runs anovas on the changes from baseline
and washout, at each visit, for each dependent variable. %Anovas
was designed to run many Anovas simultaneously. %Anovas
also reduces the pages of output created from Proc Glm by eliminating
the class page and condensing the output. This macro saves a lot of
time in the production of Anovas for many clinical research projects.
Unfortunately this macro does not allow any of the options often used
with Proc Glm.
%Anovas can be run in three modes:
Mode=1: Model dependent change from
baseline=group site covariate_list dependent_at_baseline
/ CLParm.
Mode=2: Model dependent change from baseline=group
site covariate_list dependent_at_baseline
group*site.
Mode=3: Group is continuous to test for a linear
trend. (Treatment as continous)
There should be only one observation in the dataset for each subject
visit. Visits with less then 20% of enrolled subjects are dropped.
If 80% of the data are missing, the Anova may not be informative.
Group and site should not be included the covariate list or class
list. This macro inserts group and site into the class and model statements.
Adding group or site to the Covlist will create duplicate entries
in the model statement. Site, Covlist and Classlist may be left blank.
If washout equals baseline, no washout anovas will be run. Means and
LSMeans statements, by treatment are included. Summaries of treatment
effects are printed at the end of the output. If Linear Trend(mode=3)
is selected, the summary contains the treatment effects from the type
III sum of squares from the linear trend model. Otherwise the summary
contains the treatment effects from the type III sum of squares from
the first model, without the interaction.
Syntax:
%Anovas(dataset,who,group,site,mode,visit,baseline,washout,covlist,classlist,varlist,outfile);
dataset: Dataset to use.
who: Subject
ID.
group: Treatment
group(Numeric).
site: Site(Optional).
mode: 1(Simple),
2(+treat*Site), 3(+Linear Trend).
visit: Time
variable: visit, week, month.(Numeric)
baseline: Baseline visit(Numeric
Constant).
washout: Washout visit(Numeric Constant).
covlist: Covariate
list(Optional).
classlist: Class variables(Optional).
varlist: List
of dependent variables.(Numeric)
outfile: Output file.
Example:
%Anovas(work.A,id,treatment,site,2,visit,0,0,gender
age,gender,A B C D,C:\Anova.out);
%Anovas(work.A,id,treatment,,3,visit,0,0,,,A B C
D,C:\Anova.out);
Estimate statements may be added by creating a macro named “Estimates”
before calling %Anovas:
Example:
%Macro Estimates;
Estimate 'Placebo vs Low Dose' treatment -1
1 0 0;
Estimate 'Placebo vs Med Dose' treatment -1
0 1 0;
Estimate 'Placebo vs High Dose' treatment
-1 0 0 1;
%Mend;
%Anovas(work.A,id,treatment,site,2,visit,0,0,gender
age,gender,A B C D,C:\Anova.out);
The number of digits displayed in the summary tables can be changed
by defining a macro variable named “Digits” before calling
the procedure:
%Let Digits=3;
Several styles of HTML output can be requested by defining a macro variable
named "HtmlStyle" before calling the procedure:
%Let HtmlStyle=URbw1;
Available styles include: URbw1, URpp1, BarrettsBlue, Beige, Brick, Brown,
D3D, Default, Minimal, NoFontDefault, Printer, RTF, Statdoc, Theme, FancyPrinter,
SansPrinter, SasdocPrinter and SerifPrinter.
The HTML output may be directed to a file by defining a macro variable named
"HtmlFile" before calling the procedure:
%Let HtmlFile=c:/test.html;
The size of the text in the HTML output may be changed by defining a macro
variable named "HtmlSize" before calling the procedure:
%Let HtmlSize=-1; *or;
%Let HtmlSize=+1;
The resulting output is best read and formatted
with Microsoft Word.
Output:
Output:
Output:
Please send your comments and suggestions about this web page to A. Watts
(watts@bst.rochester.edu)
| 
