I think this depends a lot on the data set and on the analysis I am doing.
So, first, look at the NUMBER of variables. If that's less than oh.... 50? 100? I'd look at means for EVERY continuous variable, unless the sample size was VERY small (say, less than 10). If there are more variables than that, then a macro that output a table
N Mean Std Min Max
for all the continuous variables would be useful, unless there were many hundreds or even thousands of variables.
>From: "Fehd, Ronald J. (CDC/CCHIS/NCPHI)" <rjf2@CDC.GOV>
>Sent: Jun 18, 2007 3:14 PM
>Subject: when do Proc Means?
>I have added a proc means/summary to my data review tool
>which I expect to publish a Beta by the end of the week.
>Since it follows a proc freq I have
>the number of rows of both the data set: NobsData
>and the proc freq output data set: NobsFreq
>Right now I have a hard-coded call to the the proc summary
>for every numeric variable,
>which returns the min, mean and max of the variable.
>Perhaps this is More Than We Wanted to Know?!
>Q: What criteria would you use to decide whether you wanted
>to see a proc means/summary of a numeric variable?
>A1: percentage: NobsFreq/NobsData gt ??%
>A2: absolute N: NobsFreq gt ??
>A?: <your criteria>
>thanx for your most(ly) expert advise on this issue
>Ron Fehd the macro maven CDC Atlanta GA USA RJF2 at cdc dot gov