John Darrington <[EMAIL PROTECTED]> writes: > On Wed, Apr 19, 2006 at 08:56:55PM -0400, Jason Stover wrote: > So now I've discovered that after I create my transformation > with add_transformation() in regression.q, I found that > close_active_file() clobbers it right after the REGRESSION > procedure is done. > > How should I avoid the call to close_active_file()? > > This is just a guess. But looking at the differences between > descriptives and regression, I see that the former has set the > transformation AFTER multipass_procedure with splits has returned.
Oh, yes, this is definitely required. Don't modify the set of transformations while a procedure is ongoing--it will not work properly, as Jason as seen. > /* Data pass. */ > ok = multipass_procedure_with_splits (calc_descriptives, dsc); > > /* Z-scoring! */ > if (ok && z_cnt) > setup_z_trns (dsc); > > Whereas regression calles add_transformation inside the > multipass_procedure_with_splits. > > I guess this is related to Ben's comments about it not working > properly with splits. Well, no, that's just because the Z-score transformation only tracks a mean and stddev for the whole file, not per split group. -- "Mon peu de succès près des femmes est toujours venu de les trop aimer." --Jean-Jacques Rousseau _______________________________________________ pspp-dev mailing list pspp-dev@gnu.org http://lists.gnu.org/mailman/listinfo/pspp-dev
