Hi:
My suggestion is that you start using the automatic percent statistics into the table one by one instead of doing them in bulk and with custom denominators.
In your ORIGINAL code, the table structure was defined like this:
table
fplgroup='FPL Group'*(agegrp='Age Group' race='Race'
firsthmpeligmonth='First HMP Elig Month'
spoken_language='Language' gender='Gender')*n='',
region='Region' all='Total';
where all the RED is the ROW dimension (the variable headers on the left side of the table) and the GREEN is the COLUMN dimension (the variable headers at the top of the table). With only 1 comma in the TABLE statement, you have a 2 dimensional table -- with only a ROW and a COLUMN dimension.
But in your changed TABLE statement , you have completely changed the structure of the table:
table
fplgroup='FPL Group'*(agegrp='Age Group' race='Race'
firsthmpeligmonth='First HMP Elig Month'
spoken_language='Language' gender='Gender')*n='',
region='Region' all='Total',
pctn<agegrp all>='Percent of row total'
pctn<race all>='Percent of row total'
pctn<firsthmpeligmonth all>='Percent of row total'
pctn<spoken_language all>='Percent of row total'
pctn<gender all>='Percent of row total'
pctn<region all>='Percent of column total'
pctn='Percent of total';
With 2 commas in the TABLE statement, you now have a 3 dimensional table with a PAGE dimension added to the table structure. Notice how your percents are shown in green, that's because when you added them after a comma in the TABLE statement, you told TABULATE to move them all into the COLUMN dimension, the REGION, ALL are being told to move into the ROW dimension and the entire list of other variables are now in the PAGE dimension (when originally, they were in the ROW dimension) -- so your new table statement, in addition to being broken, will be a completely different structure than what you originally posted.
You can only request a statistic in one dimension with TABULATE. Since you have the percents requested in the COLUMN dimension, you have to take the *n=' ' out of the PAGE dimension so that there is only one dimension (column) with a request for statistics. However, I doubt that the resulting tables will be what you want -- structure wise, at least.
As an example, run this code using SASHELP.HEART to get an idea of how the dimensions work. Notice that if I want CHOL_STATUS to ALWAYS be in the columns, with percents, then CHOL_STATUS and the statistics need to be specified after the last comma in the TABLE statement.
cynthia
proc tabulate data=sashelp.heart;
title '1) column dimension only';
class chol_status bp_status smoking_status sex;
table chol_status*(n pctn);
run;
proc tabulate data=sashelp.heart;
title '2) row and column dimension only';
class chol_status bp_status smoking_status sex;
table bp_status*sex all,
chol_status*(n pctn) all*(n pctn);
run;
proc tabulate data=sashelp.heart;
title '3) page, row and column dimension only';
class chol_status bp_status smoking_status sex;
table smoking_status all,
bp_status*sex all,
chol_status*(n pctn) all*(n pctn);
run;
