Hi, Virginia:
Generally, if I use CELLWIDTH, I do not use OUTPUTWIDTH and vice versa. Basically, OUTPUTWIDTH tells ODS PDF to go ahead and stretch (usually it's stretch) the table from margin to margin to the percent width you specify, in which case, ODS PDF wants to be in control of the CELLWIDTHS. If you specify OUTPUTWIDTH and then only SOME CELLWIDTHS, PDF could mis-calculate. I've seen instances where gaps occur, as you describe.
In addition, WIDTH and FORMAT only control width in the LISTING destination. Those particular PROC REPORT options are basically ignored by ODS PDF, RTF and HTML (along with HEADLINE, HEADSKIP, SKIP, DOL, DUL, SPACING, OL, UL, etc, etc). You may also have noticed that ODS PDF, RTF and HTML ignore Linesize and Pagesize -- although they do honor CENTER/NOCENTER.
Here are some things to try for ODS PDF:
1) take out all outputwidth and cellwidth and try this:
ODS PDF file='uniform.pdf' UNIFORM;
which attempts to determine the max variable value in every column and set the table width on every page accordingly
2) try OUTPUTWIDTH --only--
3) try CELLWIDTH --only- on both the header and column of your variable
DEFINE myvar /style(header)={cellwidth=.75in} style(column)={cellwidth=.75in};
4) use OUTPUTWIDTH & CELLWIDTH together, but make sure that you don't have an outputwidth of 100% for example, but your cellwidths add up to more or less than 100%. In this instance, I prefer to try something like this:
[pre]
options orientation=landscape topmargin=.5in bottommargin=.5in leftmargin=.5in rightmargin=.5in;
proc report data=xxx nowd
style(report)={outputwidth=10in};
. . . more code . . .
define var1 /
style(header)={cellwidth=20%}
style(column)={cellwidth=20%};
define bigvar /
style(header)={cellwidth=30%}
style(column)={cellwidth=30%};
define var2 /
style(header)={cellwidth=20%}
style(column)={cellwidth=20%};
define var3 /
style(header)={cellwidth=20%}
style(column)={cellwidth=20%};
define smallvar/
style(header)={cellwidth=10%}
style(column)={cellwidth=10%};
run;
[/pre]
On a landscape piece of paper with .5in margins on both sides, the available "real estate" is 10 inches. So I split that in percents for my 5 variables, being sure
that the cellwidth is the same for the HEADER and the COLUMN for every variable and that the total percent values add up to 100. It's probably overkill to specify the same cellwidth for HEADER and COLUMN, since I believe that the HEADER width wins, but I tend toward over-explicitness in this regard.
cynthia