summary.PSweight_sga: Summarize a PSweight_sga object

In siyunyang/PSweight.sga: PSweight.sga: Propensity Score Weighting for Causal Subgroup Analysis

Summarize a PSweight_sga object

Description

summary.PSweight_sga is used to summarize the results from PSweight_sga. The output contains the average subgroup causal effects defined by specific contrasts, as well as their standard error estimates.

Usage

## S3 method for class 'PSweight_sga'
summary(object, contrast = NULL, type = "DIF", het = FALSE, ...)

Arguments

object	a PSweight_sga object obtained from the PSweight_sga function.
contrast	a vector or matrix specifying the causal contrast of interest. The average causal effects will be defined by such contrats. For multiple treatments, the contrast parameters are explained in Li and Li (2019) for estimating general causal effects. Default is all pairwise contrasts between any two treatment groups.
type	a character specifying the target estimand. The most commonly seen additive estimand is specified by type = "DIF", abbreviated for weighted difference-in-means. This is the usual pairwise average treatment effects as those defined in Li, Morgan, and Zaslavsky (2018) and Li and Li (2019). For binary (or count outcomes), we also allow two ratio estimands: causal relative risk (type = "RR") and causal odds ratio (type = "OR"). Estimates for these two ratio estimands will be reported on the log scale (log relative risk and log odds ratio) to improve the approximate for asymptotic normality. With binary outcomes, "DIF" is the same as the average causal risk difference. Default is "DIF" if left empty.
het	an indicator specifying whether to summarize the test of heterogeneity across subgroup levels. The default is FALSE.
...	further arguments passed to or from other methods.

Details

For the contrast argument, one specifies the contrast of interest and thus defines the target estimand for comparing treatments. For example, if there are two treatment levels: A and B, the contrast A-B (i.e., E[Y(A)] - E[Y(B)]) can be specified by c(1,-1).

For estimating the causal relative risk (type = "RR"), the contrast is specified at the log scale. For example, the contrast A-B (specified by c(1,-1)) implies the estimation of log{E[Y(A)]} - log{E[Y(B)]}. For estimating the causal odds ratio, the contrast is specified at the log odds scale. For example, the contrast A-B (specified by c(1,-1)) implies the estimation of log{E[Y(A)]/E[1-Y(A)]} - log{E[Y(B)]/E[1-Y(B)]}.

The variance of the contrasts will be estimated by nonparametric bootstrap.

The argument type takes one of three options: "DIF", "RR", or "RR", with "DIF" as the default option. Typically, "RR" is relavent for binary or count outcomes, and "OR" is relavent only for binary outcomes. "DIF" applies to all types of outcomes.

Value

A list of following values:

trtgrp

a character indicating the treatment group, or target population under ATT weights.

estimates

a matrix of subgroup point estimates, standard errors and 95 for contrasts of interest.

contrast

a table listing the specified contrasts of interest.

group

a table of treatment group labels corresponding to the output point estimates, provided in results obtained from PSweight_sga.

References

Yang, S., Lorenzi, E., Papadogeorgou, G., Wojdyla, D. M., Li, F., & Thomas, L. E. (2021). Propensity score weighting for causal subgroup analysis. Statistics in medicine, 40(19), 4294-4309.

Li, F., Morgan, K. L., Zaslavsky, A. M. (2018). Balancing covariates via propensity score weighting. Journal of the American Statistical Association, 113(521), 390-400.

siyunyang/PSweight.sga documentation built on Aug. 16, 2022, 5:23 a.m.