Dr Little: Thank you for the update regarding SOLAS. Perhaps I'm misinformed, but my understanding is:
(1) SOLAS does not offer a MCMC imputation method; (2) One cannot set minimum and maximum values for variables in the imputation process (as can be done in PROC MI); (3) SOLAS cannot perform multiple imputation for logistic regression or survival models (as can be done with Harrell's transcan). Clarification of these questions/impressions would be appreciated. SR Millis At 08:48 PM 4/13/02 -0400, you wrote: >I think the latest version of SOLAS is considerably more useful, >implementing Schafer-like routines that do a much better job at >conditioning on available information. So I would not judge too much on >the early releases. Rod > >On Sat, 13 Apr 2002, Scott Millis wrote: > > > I've found SOLAS to be extremely limited--primarily using it for multiple > > imputation of regression models. Paul Alison has a paper available from > his > > website that demonstrates problems with SOLAS's propensity method. > > > > For imputation using full information maximum likelihood, AMOS works > > well. For MCMC multiple imputation, I've had good experience with SAS > PROC > > MI. Frank Harrell's SPLUS function transcan is also quite useful for a > > variety of models. Finally, the set of SPLUS functions (MICE) is worth > > looking into. > > > > SR Millis > > At 06:43 PM 4/13/02 +0300, you wrote: > > >Frank Lawrence (10.4.2002 18:33): > > > >We are contemplating a purchase of SOLAS to impute missing longitudinal > > > >data. I would like to hear from people who have used the program. > > > > > > > >Respectfully, > > > >Frank R. Lawrence > > > > > > > > > >We have used Solas for various purposes but not really for > longitudinal data. > > >Last value carried forward -method is included but it is not very much > for > > >real > > >situations. Of ourse, if a user may first work much with other tools > (like SAS > > >and own programming), and when go to Solas just for an imputation > task, the > > >results may be OK. > > >But I am interested in comments of anyone. An opportunity to discuss these > > >issues and imputation and editing questions ingeneral will be soon in > > >Jyvaskyla, > > >Finland, where an international conference, DataClean 2002, will be held, > > >29-31 > > >May. Look: > > >http://erin.mit.jyu.fi/dataclean/ > > > > > > > > > > > >___________________________________________________________________________________ >Roderick Little >Richard D. Remington Professor of Biostatistics (734) 936-1003 >U-M School of Public Health Fax: (734) 763-2215 >M4045 SPH II [email protected] >1420 Washington Hgts http://www.sph.umich.edu/~rlittle/ >Ann Arbor, MI 48109-2029
