UPHILG@UNC.BITNET (Philip Gallagher) (02/13/90)
Cindy Cummings SXA00056@BAGAMCOK writes > Is there any way to receive an output dataset with PROC TABULATE? > That is, does the OUT= option work? In passing, thanks to Prof. Bob's forbearance with respect to reading the manual. My main point: the answers to CC's question as phrased that I have read have been technically correct - i.e., no there is no OUT option or statement to produce a (SAS) dataset from PROC TABULATE. I think these answers may miss the mark, however. I have not been able to conceive of any useful meaning to the concept of a SAS dataset to be output from TABULATE (with the exception of some of the calculated statistics, which I, at least, would find easier to calculate within a DATAstep). I wonder if what CC is asking for is a way to save the formatted output from PROC TABULATE; if so, I refer her to PROC PRINTTO. That has always worked well for me. Phil Gallagher
PRAHL@MACC.WISC.EDU (Walter Prahl, MACC) (02/14/90)
Howard Schreier suggests PROC SUMMARY as an alternative to PROC TABULATE when one wants to create an output dataset. I agree, and would only add that one can often more or less duplicate the functioning of PROC TABULATE through a combination of PROC SUMMARY and PROC TRANSPOSE; SUMMARY alone will often not be sufficient. TRANSPOSE is, unfortunately, probably the most underused proc in Base SAS. People often resort to complex data step programming to do what TRANSPOSE was designed to do. Phil Gallagher suggests that an output dataset from PROC TABULATE is not a very well defined concept. While I agree that it is not entirely obvious or clear what such a thing would look like in general, I have often enough wanted to create a dataset that I could quite precisely define and that would represent something similar to what TABULATE generates. It is in these cases that I have used SUMMARY in combination with some other processing of SUMMARY's output data set (such as with TRANSPOSE). I'm not sure if one could define a reasonable dataset format that would capture all of the current variations in TABULATE's functioning, but it would certainly not be hard to define a format that would cover a large number of these variations. Basically, TABULATE would have to generate one or more auxilliary variables and place them in the dataset, like the way SUMMARY generates the _TYPE_ variable. By the way, I'm surprised that nobody has yet mentioned the fact that the addition of an OUT= option for TABULATE has been a SUGI ballot item for several years now. So this isn't exactly a new idea! ---------------------------------- Walter Prahl (608) 262-0284 University of Wisconsin -- MACC Internet: prahl@vms.macc.wisc.edu Bitnet: prahl@wiscmacc.bitnet