getjointeffects: Calculate joint effect of mixture effect and modifier vs....

In qgcompint: Quantile G-Computation Extensions for Effect Measure Modification

Calculate joint effect of mixture effect and modifier vs. common referent

Description

A standard qgcomp fit with effect measure modification only estimates effects at the referent (0) level of the modifier (psi1). This function can be used to estimate a "common referent" parameter that estimates the effect of being in a non-referent category of the modifier and increasing exposure by one quantile, relative to no change in exposure in the referent category of the modifier. This is generally useful for binary exposures (for a mixture with a set of binary exposures, this would be the "effect" of being exposed and at the index level of the mediator, relative to being unexposed in the referent level of the mediator), but it may also be of interest with more general exposures.

Usage

getjointeffects(x, emmval = 1, ...)

Arguments

x	"qgcompemmfit" object from qgcomp.emm.noboot function
emmval	numerical: value of effect measure modifier at which weights are generated
...	unused

Value

An object of class "qgcompemmeffects", which inherits from "qgcompemmfit" and "list"

This class contains the emmval-stratum specific effect estimates of the mixture. By default, this prints a coefficient table, similar to objects of type "qgcompemmfit" which displays the stratum specific joint effects from a "qgcompemmfit" model.

See Also

qgcomp.emm.noboot getstrateffects

Examples

library(qgcompint)
n = 500
dat <- data.frame(y=rbinom(n,1,0.5), cd=runif(n), pb=runif(n),
                  raceth=factor(sample(c("WNH", "BNH", "AMIND"), n, replace=TRUE),
                          levels = c("BNH", "WNH", "AMIND")))
(qfit <- qgcomp.emm.noboot(f=y ~cd + pb, emmvar="raceth",
                           expnms = c('cd', 'pb'), data=dat, q=4,
                           family=binomial()))


# first level of the stratifying variable should be the referent category,
#  which you can set with the "levels" argument to "factor" when
#  cleaning/generating data
levels(dat$raceth)

# stratum specific mixture log-odds ratios
# this one comes straight from the model (psi 1)
getjointeffects(qfit, emmval = "BNH")
# this will coincide with joint effects, since it is in the referent category
getstrateffects(qfit, emmval = "BNH")

# the stratum specific effect for a non-referent category of the EMM
#  will not coincide with the joint effect
getjointeffects(qfit, emmval = "AMIND")
getstrateffects(qfit, emmval = "AMIND")

qgcompint documentation built on March 22, 2022, 5:06 p.m.