0

I am new to imputation and EFA. I need to do EFA using 25 items on a Likert scale 1-5 or 1-4. 25% of sample have at least 1 variable with missing value. All items have some missing values. I need to use MICE to impute the missing values, but using Stata running into 'perfect predictions' issue. Getting error message even with augment option. my code below

mi set mlong
mi register imputed A...Y
mi impute chained (ologit, augment) A..Y, add (5) rseed (345)

error message

convergence not achieved
ologit failed to converge on observed data
error occurred during imputation of A...Y

Appreciate any help with solving this issue.

Nick Cox
  • 35,529
  • 6
  • 31
  • 47
bea
  • 1
  • 1
  • I’m voting to close this question because it appears to a problem with your dataset rather than a programming issue. – Nick Cox Apr 06 '21 at 10:29
  • The data has missing values, but because it is all ordinal it is tricky to impute. I can do the imputation in SAS,I also want to see if I can also do it with STATA. It is definitely a coding/syntax question. – bea Apr 06 '21 at 10:47
  • That comment doesn't make the question any easier to answer because if the data allow imputation in SAS but not in Stata, what it is about the data and/or the different syntax that makes that so? A vote to close expresses a view that the question can't be answered on the information given. I would be happy to be wrong on that. – Nick Cox Apr 06 '21 at 11:43
  • I wrote 25 separate equations for each item with fewer likely predictors with a single Proc MI statement. I reviewed STATA mi documentation, but it does not have examples addressing my situation. Meaning, how do I write 25 statements for each item using some of the other items as predictors in STATA? I am wondering if I have to create 25 subsets of data in stata for predicting each item then concatenate the data. This must be a common problem for those who do scale development given items are often in Likert scale and respondents often don't answer questions, which will cause missing values. – bea Apr 06 '21 at 22:43
  • That’s helpful detail now but I think you’d be better off taking that to Statalist or StataCorp technical support. People watching Stata here have a fair total of experience but no one with extensive -mi- experience has picked this up – Nick Cox Apr 07 '21 at 00:49
  • Thank you Nick. I will do a search to see how i could get help from them. – bea Apr 07 '21 at 21:03

0 Answers0