I have missing data values and proceeded to use multiple imputation with 138 observations. I received this error message:
Insufficient observations to estimate parameters assuming the specified prior distribution. The number of observations must be
larger than 182 for the posterior distribution to be proper. Increase the number of observations or try specifying a different prior distribution.
Am using STATA ver15.
Anyone experienced this before and what was the optimal solution?
Hi Albert,
Could you provide more details about the specific imputation model that you were using? For instance, the number and type of the variables with missing data (binary, nominal, ordinal, continuous), percentage of missing data for your variables, the MI approach that you were using (multivariate normal, chained equation), etc.
Roberto
I am using multivariate nornal and each of the 90 variables has some missing values ranging from 1 to 7.
Total observations are 101 at the moment.
As per attachment, 93% of observations have full data
I have opted for single imputation for my paper, as multiple imputation may be too advanced.
Numeric data - imputed mean
Categorical - imputed mode or median
I think that 5% missing data may not create significant bias and understatement of variance using 120+ observations.
- Badges
- Science topic