Hello,

I am analysing a dataset to recreate a prognostic score. Three of the 10 variables that go into building the score have missing data upwards of 20%. I ran multiple imputation to fill the missing values for the three variables.

However, when I use the data now to calculate the score ( varname - class; which ranges from 1-8), I get more than 1 score for the same dataset. As a result, stcox does not work because of the variation in score for the same patient.

I tried using "esampvaryok" but the results that I got were very obviously incorrect.
Code:
mi estimate, esampvaryok hr:stcox i.class , strata (trial)
I understand that averaging the scores to get the score is not the correct thing to do.
Is there anyway to have stata produce a single score for each patient in this imputed dataset?

Thank you.