@gp4 wrote: This system of styles, templates, destinations is just about incomprehensible. If you don't want to deal with a style you can provide overrides for almost any text. A brief example with defaults for the first graph and the second modifies the Colaxis values, the horizontal axis, by changing (likely) the font family and text size of the axis label and the tickmark values, plus for the label the the font style, italic, and weight, bold.   proc sgpanel data=sashelp.class; panelby sex /columns=2; scatter x=height y=weight/ group=age; run; proc sgpanel data=sashelp.class; panelby sex /columns=2; scatter x=height y=weight/ group=age; colaxis labelattrs=(Family=Arial Size=12pt Style=Italic Weight=Bold) valueattrs=(family=Serif size=6pt) ; run; With some graphs, combining Title, Footnote, Bylines, horizontal and vertical axis, legend entries, reference lines, Hline or Vline curve labels, inset statement, text plot, axistables and I'm sure few other places that you can set text attributes it could be easy to miss one in a complex graph. Once you get a custom style set to use the same font(s) in the places needed it is much easier to reference the single style than to modify the text attributes manually. And STYLES are pretty independent of the ODS Destination.   If you want to make "camera ready for publication" charts you will spend a lot of time. Getting one style that is usable with Proc SGPLOT, SGPANEL, SGPIE, SGMAP and in all the likely needed place if you are going to submit to a specific publication often is worth the time. And if you want to submit the same graphs to a different publication with different text rules then modifying a style is going to be much easier than dealing with possibly 100's of option statements in multiple procedure calls. ... View more

I have the same question. ... View more

I have been thru the documentation.  It may be a deficit in my knowledge, but seeing the definitions of the covariance estimators doesn't suggest much too me in terms of useful information.    I opted for one of the simpler ones, as the problem isn't big.  As for NLP v OPTMODEL, NLP is working and is so much less messy than OPTMODEL, that I'll stick with it for this project.    ... View more

Awesome! Great resource. Thanks again Neri ... View more

What you do depends on what exactly is the research question.  Nonetheless, to make this a typical logistic regression problem you need a yes/no outcome. You could form that by having the clinicians select a meaningful threshold, like TV reduced by 10%, yes/no.  That makes for a logistic regression, but dichotomizing data often reduces power, and sort of muddies the interpretation.  In my example, a reduction of 9.9% is a 'no' while 10.00 is a 'yes', but is there a meaningful difference in these two volumes?   Relative % change = (TV1-TV3)/TV1 = 1-TV3/TV1.  Even if TV is normally distributed the ratio TV3/TV1 almost surely isn't normal.  From your data example it doesn't look like TV2 is used.   So you could use the Kruskal-Wallis test, available with the Wilcoxon option in proc npar1way, to compare groups and not have to be concerned with parametric assumptions.   If the overall test is significant, the DSCF for the pairwise comparisons.   There is a literature, mostly based on simulations, suggesting that ANCOVA with TV1 as the covariate is the preferred analysis.  Here is an easy read on the subject:  https://www.ncbi.nlm.nih.gov/pmc/articles/PMC34605/       ... View more

Now that I know I feel like I should have seen that!  Thank you.   ... View more