Hi all,

I am looking to construct an index (ranging from 0 to 1) describing how positively the survey respondents evaluated a certain poll. The index is made from the mean of three indicator variables: informative, accurate, and trustworthy, all addressing different aspects of poll evaluation. Given then presence of missing values across these three indicators, I have tried imputing these, using the code below:

mi impute chained (truncreg, ll(0) ul(1)) informative accurate trustworthy (ologit) education = age female, add(20) rseed(2232)

Now, I plan to use - mi estimate - as a next step, with the poll evaluation index acting as the dependent variable, however, I'm not sure how I can go about constructing the index? Is it a case of using - mi passive- to generate the mean from each of the imputed variables ? When I use the following syntax

mi passive: egen pollevaluation = rowmean(informative accurate trustworthy)

I simply get the mean of the variables in m=0.

Any help much appreciated!

Leo