Mi estimate: command not supported

[manu6msa]

I am using as a dependant variable the Gini index. It comes from the Standardized World Income Inequality datset. This variable is based on imputations from different income inequality dataset. Therefore i need to use the command "mi estimate" for all my estimations. I am working with a panel data of countries. The problem is that "mi estimate" does not support the command "xtivreg", which i need to do an instrumental variable estimation.
According with STATA manual the option [,cmdok] allows estimation when the command is not supported by "mi estimate". However I have been trying to find how this option works but I have not been able to do it. So my question is if it is possible to use the instrumental variable estimation with "mi estimate" although the command "xtivreg" is not supported by mi estimate.

Thanks in advance for any help.

[Stephen Jenkins]

Welcome to Statalist! According to [MI] Manual entry for mi estimate, commands that are not on the list of commands given there are not supported. User-written commands (such as xtivreg) can be made to work with mi estimate if property mi is added to the program's properties, and [P] Program Properties is cross-referenced. I take this to mean that you'll have to contact the authors of xtivreg and ask them.
By the way, please may I suggest that you read the Forum FAQ (hit the black bar at the top of the page) before sending your next post. You'd maximise your chances of getting a helpful response. Contrary to FAQ recommendations, you haven't explained the provenance of user-written command xtivreg; you haven't shown us exactly what you typed and exactly what Stata returned; you've misspelt "Stata" (it's not "STATA"); and you haven't registered with your real name (firstname familyname). To re-register, I suggest that you hit the Contact Us button on the bar at the bottom of the screen and ask the Forum administrators to change your login name (they are fast and efficient). Thank you.

[Richard Williams]

See this thread, especially Yulia M's posts and links: http://www.statalist.org/forums/foru...itten-commands

It would be great if the authors of xtivreg could tell you why they did or did not support MI. However, the answer may be that they didn't think about it or don't know if it is ok or not. My own feeling is that, if it runs and the estimates seem reasonable, go ahead and use cmdok. However, I would also check to see if multiple imputation is gaining me that much over listwise deletion. Often it doesn't.

[Phil Schumm]

I am using as a dependant variable the Gini index. It comes from the Standardized World Income Inequality datset. This variable is based on imputations from different income inequality dataset. Therefore i need to use the command "mi estimate" for all my estimations.

Quick question: Do you actually have multiple imputations for each observation, or do you only have one value of the Gini index for each observation (even though that value may have been obtained through a process involving multiple imputation)? If the latter, then use of mi estimate is not appropriate, regardless of which estimation command you are using. I apologize if this question is unwarranted, but from what you've written, I can't be sure.

[Jonathan Horowitz]

I'm wondering, did anyone figure out if it's okay to use mi estimate with xtivreg? After looking through things I can't find anything showing it wouldn't be fine, but if someone has something more definitive I'd love to know.

See this thread, especially Yulia M's posts and links: http://www.statalist.org/forums/foru...itten-commands

It would be great if the authors of xtivreg could tell you why they did or did not support MI. However, the answer may be that they didn't think about it or don't know if it is ok or not. My own feeling is that, if it runs and the estimates seem reasonable, go ahead and use cmdok. However, I would also check to see if multiple imputation is gaining me that much over listwise deletion. Often it doesn't.