2
$\begingroup$

I am working with a database with missing data. I have done "Roderick J. A. Little’s chi-square statistic" and knew that my data are not MCAR. However, I know don't have can I determine if there are MAR or NMAR. I was wondering if there is a test that you may suggest me. Also, how we can know that how many imputations we have to run and what factors determine the number of imputation we should apply.

$\endgroup$
1
  • $\begingroup$ Can you try to clarify this question? Title seems to be divergent from the contents. (@MTibbits Unfortunately nothing better than asking the OP do it.) $\endgroup$ – user88 Jun 29 '11 at 10:29
1
$\begingroup$

There is no formal test for MAR vs. NMAR, as far as I know. Point estimates are generally believed to be well estimated in most situations with a small number of imputations, but standard errors and confidence intervals require somewhat more. If computing resources are available to do so, using more is better. Here's a good place to look for advice: https://stefvanbuuren.name/fimd/sec-howmany.html.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.