I have it, but currently only in PL/1 (mainframe) and perl. I also have documentation from 1980 on the method but I would have to dig for it. Let me know if you can't get Excel code and want what I have. However, I discovered to my distress that the Bureau has changed the rules on us for 2000. In 1980 and 1990 medians were computed from the grouped data as reported in STF3. This means that 17 intervals were used in 1980, 25 in 1990. When I couldn't get medians that matched the Bureau's for the General Demographic Profiles, I guessed it was because there are fewer intervals than there will be in the full SF3. I contacted the Bureau and found out that the medians are calculated "behind the scenes" from 39 intervals, although only 17 intervals will be in SF3 (according to prelim doc, and the numbers 17 and 39 are from memory--they are close if not exact). So this means we will not be able to exactly replicated Census methodology. It also means we can't calculate standard errors for medians. I calculated the median income for Cook County from the GDP using Pareto and it is about $500 different than the Bureau's number. It should be closer when we get the real SF3. I posted the Bureau's response about this on the State Data Center Listserve. If anyone wants it I can send them a copy or post here. Jim Bash On Tue, 18 Jun 2002, Daryl Scott wrote: