Roy,

Another possibility you might take a look at is the macro which York University's (Toronto) Michael Friendly has provided for disaggregating data http://www.math.yorku.ca/SCS/sasmac/stat2dat.html

It takes aggregated data, likes n, mean, stdev, etc. and converts it into an analyzable dataset. I would think that the disaggregated data sets could be easily combined and reanalyzed.

Art ----------- On Thu, 26 Jan 2006 13:14:31 -0800, Pardee, Roy <pardee.r@GHC.ORG> wrote:

>Hey All, > >I'm getting some proc tabulate output from various sites, which will >contain the statistics N Min Max Mean & Std. I'm tasked with collating >these stats into a single uber-table, which must show what we would get >if I had access to all the raw data available at the various sites. > >Those first four stats are reasonably easy to combine (add the N's, take >min() of the Min's, use the N's to calculate a weighted mean, etc.) but >the Std has me stumped. > >Is there some proc out there that will take something like this for an >input dset: > >Obs MAX MEAN MIN N STD site > > 1 5 4 3 4 0.8165 A > 2 65 34 3 2 43.8406 B > >And give me what the STD of the union of the samples would be? > >Thanks! > >-Roy