I just noticed a peculiar behavior that I thought I'd share.
If I do this...
gen byte temp1 = V8==.c
egen byte b1 = max(V8==.c), by(V2 V3)
egen byte b2 = max(temp1), by(V2 V3)
Then b1 and b2 can be different, contrary to my initial expectation.
This was a puzzle, but then I realized that egen.ado (and _gmax.ado) set
version to 6. Therefore, the expression V8==.c within the call to egen
evaluates to 1 for ALL missings -- not just .c. Thus, b1 indicates the
presence of any missings in V8, whereas b2 indicates the presence of .c only.
This is something to be aware of, and I didn't see any warning about it in
the egen help.
I''ve sent this to Stata tech support, too.
Institute for Policy Studies
Johns Hopkins University