Stata's new merge is intriguing and perhaps worrisome.
Consider this:
merge patkey newenc using ${sql}demo1 ${sql}surv3,keep(hlth* w_hlthins)
to which Stata replies: "variable w_hlthins not found in using dataset"
Done the old way:
merge patkey newenc using ${sql}demo1,keep(hlth*)
merge patkey newenc using ${sql}surv3,keep(w_hlthins)
Stata has no problem. Therefore the -merge- change for multiple merges
using a -keep()- option only works if the 'kept' variables are in both data
sets.
At the very least Stata might upgrade the help file to make this manifest.
Of course, Stata might look ahead and see if the 'kept' variable was
present in any of the using data sets.
Also, _merge/k refers to the relation between master and using data
sets(k). However, what if the first merge alters the master set such that
the new master yields _merge results with respect to using set k+1.
There is case to be made for merging on set at a time. Hard to see that
multiple merges add much utility for the potential confusion.
Fred Wolfe
Fred Wolfe
National Data Bank for Rheumatic Diseases
Wichita, Kansas
Tel (316) 263-2125 Fax (316) 263-0761
fwolfe@arthritis-research.org
