Impute using multivariate normal regression (mi impute mvn)

[Carolina Hincapie]

Dear Statalist community,

I am seeking guidance as I am facing challenges in performing multiple imputation using the mi impute mvn command (using Stata 18).

Here are the steps I have taken:

• Loaded the example dataset (I prefer to use first example data bases for learning before going to perform it in my real database):

use https://www.stata-press.com/data/r18/mheart5s0.dta

• Created a table to examine missing data:

mi misstable summarize, all

• Realized that I needed to impute data for the variables age and bmi.
• Attempted to impute multiple variables simultaneously:

mi impute mvn bmi attack smokes age female, add(20) rseed(1234)

• Faced an issue with variables attack, smokes, and female not being registered as imputed; hence, unregistered and registered them:

mi unregister attack smokes female mi register imputed attack smokes female

• Tried the imputation command again:

mi impute mvn bmi attack smokes age female, add(20) rseed(1234)

• Received this output:
  note: variables attack smokes female contain no soft missing (.) values; imputing nothing
  
  Performing EM optimization:
  observed log likelihood = -530.87093 at iteration 11
  
  Performing MCMC data augmentation ...
  
  Multivariate imputation Imputations = 40
  Multivariate normal regression added = 20
  Imputed: m=21 through m=40 updated = 0
  
  Prior: uniform Iterations = 2000
  burn-in = 100
  between = 100
  
  ------------------------------------------------------------------
  | Observations per m
  |----------------------------------------------
  Variable | Complete Incomplete Imputed | Total
  -------------------+-----------------------------------+----------
  bmi | 126 28 28 | 154
  attack | 154 0 0 | 154
  smokes | 154 0 0 | 154
  age | 142 12 12 | 154
  female | 154 0 0 | 154
  ------------------------------------------------------------------
  (Complete + Incomplete = Total; Imputed is the minimum across m
  of the number of filled-in observations.)
• Saved the imputation results:

save final_imputation_data

• Checked the imputed data using edit or mi misstable summarize, all, but still found the same missing values.

I am uncertain about what I might be doing wrong and how to save the imputed data correctly.
I already try to have a look on mimiimputemvn.pdf (stata.com) and through previous post, but couldn't find an answer.
Any assistance from the Statalist community would be greatly appreciated.

Sincerely,

Carolina

[daniel klein]

Have a look at [MI] Styles -- Dataset styles (p. 352--359) to understand how imputed, i.e., completed, datasets are stored. The example data you use is in mlong style. You can find that out by typing

mi query

The observations with missing values are still in the same dataset after imputation. There is nothing wrong with that.

[daniel klein]

By the way, conceptually, you probably want to list the regular variables as predictors in your imputation model, like this:

mi impute mvn bmi age = i.attack i.smokes i.female i.hsgrad , add(#) ...