Re: st: Missing Values in Multiple Imputation Data Sets

[Richard Williams <richardwilliams.ndu@gmail.com>]

One easy thing to check: Stata has both hard and 
soft missings. Soft missings (coded .) can be 
imputed, hard missings (coded .a through .z) 
cannot. So check to see if you are using hard 
missings, and if so determine whether it is 
reasonable to recode them as soft missings. You 
may not want to recode if the hard missings 
reflect things like Not Applicable responses.

Beyond that the messages are pretty clear. You 
can increase the matsize, make sure you aren't 
treating a continuous variable as categorical, etc.

Especially if you have only been using Stata for 
3 weeks, I would recommend starting simple and 
building up. This looks like a pretty complicated 
model. Start with fewer variables and build up. 
It may be easier to identify which variables are causing you grief.

At 06:55 AM 2/5/2013, Patti Fritz wrote:
> Hello All,
>
> I am a novice to Stata, having just begun using 
> it 3 weeks ago. Therefore, my questions are 
> likely very naive. I am trying to conduct 
> multiple imputation of a data set that primarily 
> contains positively skewed/frequently 
> over-dispersed ordinal predictor variables 
> (i.e., history of various forms of 
> family-of-origin aggression assessed on 3- or 
> 4-point scales from not at all to a lot); my 
> other predictors consist of demographic 
> variables (e.g., age, gender, income, number of 
> marriages, marital status). My outcome variables 
> are dichotomous (partner violence) and the 
> sample size is N = 3,635. There is substantial 
> missing data on some of the items (e.g., 21%), 
> and complete data across all variables is only 
> found among 556 cases). When I attempt to 
> conduct mi impute chained commands (I've tried 
> poisson, and nbreg), there are still some 
> missing values in some of the imputed data sets. 
> Specifically, every variable still has some 
> missing data across the 20 imputed data sets. Is there!
>   a way to alleviate this problem?
>
> I can get the mi impute mvn command to work, but 
> after reading some of the literature, I'm not 
> sure this is the best approach. In addition, I 
> haven't been able to get the mi chained (ologit) 
> command to work because I get the following error:
>
> matsize too small
> Â Â Â  You have attempted to create a matrix 
> with too many rows or columns or attempted to fit a
>     model with too many variables.  You need 
> to increase matsize; it is currently 800.  Use set
> Â Â Â  matsize; see help matsize.
>
> Â Â Â  If you are using factor variables and 
> included an interaction that has lots of missing
> Â Â Â  cells, either increase matsize or set 
> emptycells drop to reduce the required matrix size;
> Â Â Â  see help set emptycells.
>
> Â Â Â  If you are using factor variables, you 
> might have accidentally treated a continuous variable
> Â Â Â  as a categorical, resulting in lots of 
> categories.  Use the c. operator on such variables.
> error occurred during imputation of g2sabuse_s05 ipvabuse_s05 s2gabuse
> _g04 ipvabuse_g04
> f2sslap_s94 m2sslap_s94 f2gslap_s94 m2gslap_s94 
> s2gslap_s94 g2sslap_s94 p2pslap_s94 f2sslap_s05
> f2sabuse_s05 m2sslap_s05 m2sabuse_s05 
> f2gslap_g04 f2gabuse_g04 cp260hec cb103redr gb103redr on m
> = 1
> r(908);
>
> I only have Stata/IC 12.1 on my laptop, but I'm 
> going to see if the computer labs at work have 
> the SE version. If yes, I'll see if I can get 
> the ologit command to work. Regardless, I'm 
> interested in finding out why missing values are 
> found in the imputed data sets.
>
> Thanks in advance for your time and assistance. They are greatly appreciated!
>
> Cheers,
>
> Patti
>
>
>
> Patti A. Timmons Fritz, Ph.D., C. Psych.
> Associate Professor
> Department of Psychology
> University of Windsor
> Windsor, Ontario N9B 3P4
> Canada
>
>
>
> *
> *   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/

-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam


*
*   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/