Validity of reporting variance ratios for binary covariates

[Hannah Beilby]

Hi folks,

Wondering why tebalance summarize allows variance ratios to be shown, when the R "matchit" package refuses, with justification:

The variance ratios are computed as the ratio of the variance of the treated group to that of the control group for each covariate. Variance ratios are not computed for binary covariates because they are a function of the prevalence in each group, which is captured in the mean difference and eCDF statistics.

See:

 https://cran.r-project.org/web/packages/MatchIt/vignettes/assessing-balance.html

All of my variables are binary, therefore think there is no extra value add for reporting the variance ratios if I am already displaying the standardised differences before and after inverse probability weighting?

Any reassurance or other opinions gratefully received!

Kind regards,
Hannah

[Daniel Schaefer]

This really feels like a question for a mentor or advisor in your field.

Yes, the R documentation seems to suggest the variance ratio's for dichotomous variables are redundant. That said, why not report them in a table anyway? You might try to think about what the common practice is in your field. If people in your field generally understand that variance ratio's for dichotomous variables are redundant then there is no need to report them. If this is not common knowledge, you could chose to include or exclude them, but if you choose to exclude them you should probably explain why.

This post is duplicated here. Sometimes top level posts take a while to finish submitting. Post duplications sometimes happen when you spam click the "post" button. You can avoid this problem by simply waiting for the page to finish reloading. It might take a few seconds.