psweights: Propensity Score Weights

In dewittpe/pstools: Propensity Score Analysis Tools

Description

Five different matching weights based on an exposure and propensity score.

Usage

psweights(.data, exposure, ps)

psweights_(.data, exposure_col, ps_col)

## S3 method for class 'pstools_psweights'
plot(x, ...)

Arguments

.data	a data.frame
exposure	the bare name for the column within .data indicating the exposure or non-exposure. Expect this to be an integer, or coercible to an integer with values of zero or one.
ps	the propensity scores. Expected values between 0 and 1.
exposure_col	a character string
ps_col	a character string
x	a pstools_psweights object
...	ignored

Details

Let p denote the propensity score for a subject. The five weights are as follows.

		Exposed	Non-Exposed
Average causal effect in study population	psw_IPW	1/p	1/(1-p)
Average causal effect in exposed	psw_ACE_Exposed	1	p/(1-p)
Average causal effect in unexposed	psw_ACE_Unexposed	(1-p)/p	1
Average causes effect in population for which sample is most informative	psw_ACE_MostInfo	1-p	p
Average causal effect in mathcin weight population	psw_ACE_MWP	min(p,(1-p))/p	min(p,(1-p))/(1-p)

Value

a data.frame with the exposure and ps vectors returned along with five different weights. See Details for information on the five weights.

Methods (by generic)

• plot: Mirrored histograms

References

Li, Liang, and Tom Greene. "A weighting analogue to pair matching in propensity score analysis." The international journal of biostatistics 9.2 (2013): 215-234.

Examples

glmfit <- stats::glm(PCR_RSV ~ SEX + RSVINF + REGION + AGE + ELTATOP + EINZ + EXT, 
                     data = pride,
                     family = stats::binomial()) 

ourpride <- dplyr::mutate(pride, pp = qwraps2::invlogit(fitted(glmfit))) 

psw <- psweights(dplyr::tbl_df(ourpride), PCR_RSV, pp)
psw

# Mirrored Histograms
plot(psw)

dewittpe/pstools documentation built on May 15, 2019, 5:07 a.m.