I have a few questions regarding multiple imputation for nested data. Context: I have repeated measures (4 times) from a survey and these are clustered in workplaces (205 workplaces). There are about 180 items on this survey.

q1. Is it possible to take both the repeated measures and the workplace clustering into consideration or do i have to decide for one of the two?

q2. If i can only take into consideration one of the two clusterings (repeated measures vs workplace) which one would you recommend

q3. I have about 10000 observations and about 400 of them have missing values for the workplace. What would you recommend to do in this case? (also i should mention that the 205 workplaces are nested in 17 Organizations - For the moment i use general categories based on the organization: e.g. Organization1-Unclassified). Is there a meaningful way to actually impute these categories?

q4. Would you recommend to use all 180 items for imputation or the items that i intend to use in each of my models?

I use R for analysis and it would be greatly appreciated if you can recommend any packages for multiple imputation for clustered data.

Thanks in advance


1 Answer 1


Take a look at the Amelia II package, by Honaker, King and Blackwell.

Amelia II "multiply imputes" missing data in a single cross-section (such as a survey), from a time series (like variables collected for each year in a country), or from a time-series-cross-sectional data set (such as collected by years for each of several countries).

q1. Yes, it is possible with the above package.

q3. I guess it would be possible to impute workplace using a more general imputation method. (MICE could help).

q4. As a general rule, do not throw information out. The imputation model, at a minimun, should include all covariates in all your models. But if extra information help predicting the missing data, include it!


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.