Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Starting values in -mi impute mlogit-
From
[email protected] (Yulia Marchenko, StataCorp)
To
[email protected]
Subject
Re: st: Starting values in -mi impute mlogit-
Date
Fri, 21 Jun 2013 12:03:24 -0500
Brendan Halpin <[email protected]> tried to use -mlogit-'s -from()- option
with -mi impute mlogit- and received an error that this option is not
supported, although it is documented:
> I am running multiple imputation (-mi impute mlogit-) where I have occasion
> to re-run an imputation on data that is very little changed. To save time I
> have tried to save the estimates and feed them to the second invocation as
> starting values. The help suggests (not very explicitly) that this should be
> possible: under -help mi impute mlogit- the -maximize_options- header
> points to -help mi mlogit- which says the -from()- option can be used. While
> this is correct for plain -mlogit-, -mi impute mlogit- complains:
>
> . mi impute mlogit state i.last i.next, add(1) augment force from(b, skip)
> option from() not allowed
>
> So, is there a way to pass starting values to -mi impute mlogit-?
Brendan is using -mi impute-'s option -augment- to handle perfect prediction
during multinomial-logit imputation. For speed purposes, the augmentation
procedure uses an internal command, -_mlogit-, which does not support the
-from()- option.
We will look into the possibility of accommodating the specification of
initial values during the augmentation procedure. If this proves infeasible,
we will update our documentation to clarify which of -mlogit-'s options are
supported when the -augment- option is specified.
-- Yulia
[email protected]
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/