R: Plotting Covariate Balancing Propensity Score Estimation

plot.CBPS {CBPS}

R Documentation

Plotting Covariate Balancing Propensity Score Estimation

Description

This function plots the absolute difference in standardized means before and after weighting. To access more sophisticated graphics for assessing covariate balance, consider using Noah Greifer's cobalt package.

Usage

## S3 method for class 'CBPS'
plot(x, covars = NULL, silent = TRUE, boxplot = FALSE, ...)

Arguments

`x`	an object of class “CBPS” or “npCBPS”, usually, a result of a call to `CBPS` or `npCBPS`.
`covars`	Indices of the covariates to be plotted (excluding the intercept). For example, if only the first two covariates from `balance` are desired, set `covars` to 1:2. The default is `NULL`, which plots all covariates.
`silent`	If set to `FALSE`, returns the imbalances used to construct the plot. Default is `TRUE`, which returns nothing.
`boxplot`	If set to `TRUE`, returns a boxplot summarizing the imbalance on the covariates instead of a point for each covariate. Useful if there are many covariates.
`...`	Additional arguments to be passed to plot.

Details

The "Before Weighting" plot gives the balance before weighting, and the "After Weighting" plot gives the balance after weighting.

### @aliases plot.CBPS plot.npCBPS

Value

For binary and multi-valued treatments, plots the absolute difference in standardized means by contrast for all covariates before and after weighting. This quantity for a single covariate and a given pair of treatment conditions is given by \frac{\sum_{i=1}^{n} w_i * (T_i == 1) * X_i}{\sum_{i=1}^{n} (T_i == 1) * w_i} - \frac{\sum_{i=1}^{n} w_i * (T_i == 0) * X_i}{\sum_{i=1}^{n} (T_i == 0) * w_i}. For continuous treatments, plots the weighted absolute Pearson correlation between the treatment and each covariate. See https://en.wikipedia.org/wiki/Pearson_product-moment_correlation_coefficient#Weighted_correlation_coefficient.

Author(s)

Christian Fong, Marc Ratkovic, and Kosuke Imai.