No announcement yet.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Multiple imputations and descriptive statistics

    So, I just found out that using tab when you have multiple imputations is a big no no. This was a bit of a bummer since I didn't know this and had multiple 2 by 2 tables as descriptive statistics. So now I need to redo that.

    I am just wondering how.

    So I have a variable that is binary, I just want to see how many said yes or no and the % split. I used fre for that, I assume this also won't work anymore with MI? For that I will use mi estimate: total I guess and just calculate the % split, doesn't seem to have an option to include that in the output.

    I also have tables with 2 variables that I need to redo. On one side are ages split into groups (16-25, 26-35 etc.) and on the other the same with monetary values (0-10k 10k -100k etc.)

    One way could be

    forvalues j=0/5{
      tab age money if implicate==`j', cell nofreq
    This should work with multiple imputations right? But how do I get a table with all imptuations combined into one as tab (incorrectly) would have done? Is this where I use the "over" option for mi estimate: total?

    Or do you have any other recommendations of what ways I could create simple descriptive tables when dealing with multiple imputations?

  • #2
    a very useful community-contributed command for calculating descriptive statistics after -mi- is -misum- by daniel klein (-search misum-).
    Kind regards,
    (Stata 16.0 SE)


    • #3
      Carlos as far as I can tell mi sum functions similarly to sum. It gives me a mean, SD, min and max value. However, how do I get a descriptive table where I combine 2 variables in the same way tab allowed me?

      I attached an image of what I was hoping to do with my data. Is this possible even with MIs?

      Click image for larger version

Name:	age and gift class.png
Views:	1
Size:	6.7 KB
ID:	1499416


      • #4
        to the best of my knowledge, I do not remember user-written commands for -mi- that can accomplish that.
        Kind regards,
        (Stata 16.0 SE)