Both Enzo Coviello <[email protected]> and poker buddy Paul Dickman
<[email protected]> have had trouble with predict after using -glm- with
a user-defined relative survival link.
The cause of their problems is a bug introduced in Stata 9. The bug
affected cases where the global macro SGLM_p, used by -glm-, was not being
handled properly by -predict- when the macro itself was non-numeric. This
only affects cases where this macro is being used for non-standard purposes,
for instance, to hold a variable name for population survival rates.
The fix to this bug will be available in a future ado update.
