Announcement

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

  • Using svy and mi prefixes: How to complete a descriptive table on svy subpopulations after multiple imputation?

    Hi all,

    I am working with nine waves of a large, nationally-representative cross-sectional dataset with information on multiple dimensions of population health and determinants (NHANES). I chose to conduct multiple imputation to preserve observations with missing information on some covariates. Now that I have completed this, I am having issues running the commands to complete my descriptive table.

    For example, my dataset includes five racial/ethnic subpopulations, and I would like to obtain proportions (and ideally observation counts for each cell, along with a chi-square stat for difference) on all of the covariates across these five subpopulations.

    I am attempting to do so in the following way with:

    mi estimate: svy, subpop(if studypops2==1): proportion age_cat
    mi estimate: svy, subpop(if studypops2==2): proportion age_cat
    mi estimate: svy, subpop(if studypops2==3): proportion age_cat
    mi estimate: svy, subpop(if studypops2==4): proportion age_cat
    mi estimate: svy, subpop(if studypops2==5): proportion age_cat

    I am running into some issues with this: 1) there is no option to obtain the number of observations, 2) there is no option to obtain a chi-square stat for difference. These issues, while inconvenient, are not as pressing as the fact that the number of observations in the population is changing depending on which subpop I am focusing on. The note given by Stata was:

    "Note: 26 strata omitted because they contain no subpopulation members."

    Does anyone know of 1) a better way to execute the descriptive statistics with svy data that has been multiply imputed? Or, 2) any tweaks to make my current approach work better/prevent strata from being omitted?

    Many thanks,
    Maria


  • #2
    Hardt recommends to report descriptive statistics before the multiple imputation process, with missing %s for variables so readers can see the original data and assess potential bias.

    Hardt, J., Herke, M., Brian, T., & Laubach, W. (2013). Multiple Imputation of Missing Data: A Simulation Study on a Binary Response. Open Journal of Statistics, 03(05), 370–378. https://doi.org/10.4236/ojs.2013.35043

    Comment

    Working...
    X