Dear Richard,
- making the impute.ado adjustment as you suggested sounded like a logical step. I did it, but it would refuse to recognise my newly created impute2.ado for some reason, though I saved it in the same folder as where impute.ado was located. By the way, looking at the ado file, I now see why -set type double, permanently- doesn't do the trick... because the ado file asks for new vars to be created as float, and the -set- just gives guidelines for new vars for which the type is not explicitly specified.
- unfortunately I do need that level of precision, since my variables have large values. So -recast- would not undo the discrepancies due to -impute-'s use of float types.
- I too am surprised about the forced 'float' in -impute-. My email you were responding too unfortunately didn't chain to the email trail on this issue in statalist despite the use of the same subject line, so here is where you can find my demonstration of how -impute- force-uses float (and the concomitant loss of precision in case of large values is obvious and need not be demonstrated):
http://www.stata.com/statalist/archive/2008-07/msg00519.html
Hewan
---------------------------------------------------------------------------
