Maximum Pairwise Standardized Mean Difference

Usage

max_pairwise_smd(x, name, round_to = 3, ...)

Arguments

x: named (list)
Where each list element is a vector corresponding to the observations of the variable of interest for that particular strata (corresponding to the name). See example for how this can be constructed.
name: (character)
Unused variable that is required to be supported by extra.col in table1::table1().
round_to: (numeric)
How many digits to round the standardized mean difference to.
...: Additional arguments passed on to table1::table1().

Value

(numeric)
The maximum pairwise standardized mean difference between all strata for a particular variable.

Note

The implementation with the stddiff package is more fragile, than with the smd package. However, the smd package uses the population variance to calculate the SMD as opposed to the sample variance. This can cause small inaccuracies in the final result. Therefore we elect to implement with stddiff. Another consideration is that stddiff can be maximally precise to only 3 decimal places.

Additionally, for very large cohorts, standardized mean differences calculated on categorical variables using the stddiff package may throw In n[1] * n[2] : NAs produced by integer overflow. This is an implementation issue in stddiff in the calculation of the standard errors for the standardized mean differences. However, since these standard errors are not used in the final output, they can be safely ignored.

Examples

max_pairwise_smd(split(mtcars$disp, mtcars$am))
#> [1] 1.478