Re: st: st: Using MVN for Multiple Missing Ordinal Variables

[Nick Cox <njcoxstata@gmail.com>]

I would reverse this question. On what grounds could multivariate
normal possibly seem right for ordinal data? Why are you considering
such distributions at all? Presumably you have some information on the
distributions of your ordinal variables. If they seem like
integer-rounded variants of normals, then MVN might be the best thing
you can come up with, but it hardly seems appropriate or attractive
otherwise.

Likert scales are named for Rensis Likert, so capitalisation is recommended.

Nick

On Sun, May 8, 2011 at 2:15 PM, Clifton Chow
<clifton_chow@post.harvard.edu> wrote:
> I have multiple missing values from a survey administered to a sample of just slightly over 150 persons.  The items are structured in a ordered likert scale of between 5 and 8 items.  I have checked missing patterns with misstable and most are not monotone.  I need to impute missing values and generate descriptive statistics.  I am wondering if Impute MI MVN method (mutivariate normal) with stata 11 would be appropriate for ordinal data?  If it is, once I generate the 10 imputed datasets, should I simply average the imputed values and adjust the variance?  Does anyone with experience working with missing values have any other suggestion for ordinal data?

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/