qgcompint (version 1.0.2)

qgcomp.emm.glm.ee: EMM for Quantile g-computation for continuous, binary, and count outcomes under linearity/additivity

Description

This function estimates a dose-response parameter representing a one quantile increase in a set of exposures of interest at levels of a binary, factor, or continuous covariate. This allows testing of statistical interaction as well as estimation of stratum specific effects. This model estimates the parameters of a marginal structural model (MSM) based on g-computation with quantized exposures. Note: this function allows clustering of data and/or sampling weights and yields cluster-robust standard errors for all estimates

Estimating equation methodology is used as the underlying estimation scheme. This allows that observations can be correlated, and is fundementally identical to some implementations of "generalized estimating equations" or GEE. Thus, it allows for a more general set of longitudinal data structures than does the qgcomp.glm.noboot function, because it allows that the outcome can vary over time within an individual. Interpretation of parameters is similar to that of a GEE: this function yields population average estimates of the effect of exposure over time.

Usage

qgcomp.emm.glm.ee(
  f,
  data,
  expnms = NULL,
  emmvar = NULL,
  q = 4,
  breaks = NULL,
  id = NULL,
  weights,
  offset = NULL,
  alpha = 0.05,
  rr = TRUE,
  degree = 1,
  includeX = TRUE,
  verbose = TRUE,
  errcheck = TRUE,
  ...
)
qgcomp.emm.ee(
  f,
  data,
  expnms = NULL,
  emmvar = NULL,
  q = 4,
  breaks = NULL,
  id = NULL,
  weights,
  offset = NULL,
  alpha = 0.05,
  rr = TRUE,
  degree = 1,
  includeX = TRUE,
  verbose = TRUE,
  errcheck = TRUE,
  ...
)

Value

a eeqgcompfit/qgcompemmfit/qgcompfit/list object, which contains information about the effect measure of interest (psi) and associated variance (var.psi), as well as information on the conditional/underlying model fit (fit) and the marginal structural model fit (msmfit).

Arguments

f

R style formula

data

data frame

expnms

character vector of exposures of interest

emmvar

(character) name of effect measure modifier in dataset (if categorical, must be coded as a factor variable)

q

NULL or number of quantiles used to create quantile indicator variables representing the exposure variables. If NULL, then gcomp proceeds with un-transformed version of exposures in the input datasets (useful if data are already transformed, or for performing standard g-computation)

breaks

(optional) NULL, or a list of (equal length) numeric vectors that characterize the minimum value of each category for which to break up the variables named in expnms. This is an alternative to using 'q' to define cutpoints.

id

(optional) NULL, or variable name indexing individual units of observation (only needed if analyzing data with multiple observations per id/cluster). Note that qgcomp.noboot will not produce cluster-appropriate standard errors (this parameter is essentially ignored in qgcomp.noboot). Qgcomp.boot can be used for this, which will use bootstrap sampling of clusters/individuals to estimate cluster-appropriate standard errors via bootstrapping.

weights

"case weights" or sampling weights - a vector of weights representing the contribution of each observation to the overall fit

offset

Model offset term on individual basis: not yet implemented glm or bayesglm

alpha

alpha level for confidence limit calculation

rr

(logical, default=TRUE) estimate a risk ratio from the MSM when using an underlying logistic model

degree

polynomial degree for non-linearity of MSM (values other than 1 are not supported, currently)

includeX

(logical, default=TRUE) include design matrixes in the output?

verbose

(logical, default=TRUE) include informative messages

errcheck

(logical, default=TRUE) include some basic error checking. Slightly faster if set to false (but be sure you understand risks)

...

arguments to glm (e.g. family)

See Also

qgcomp.noboot

Examples

set.seed(50)
# linear model, binary modifier
dat <- data.frame(y=runif(50), x1=runif(50), x2=runif(50),
  z=rbinom(50, 1, 0.5), r=rbinom(50, 1, 0.5))
(qfit <- qgcomp.emm.glm.noboot(f=y ~ z + x1 + x2, emmvar="z",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
(qfitee <- qgcomp.emm.glm.ee(f=y ~ z + x1 + x2, emmvar="z",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
# logistic model, continuous modifier
dat2 <- data.frame(y=rbinom(50, 1, 0.5), x1=runif(50), x2=runif(50),
  z=runif(50), r=rbinom(50, 1, 0.5))
(qfit2 <- qgcomp.emm.glm.noboot(f=y ~ z + x1 + x2, emmvar="z",
  expnms = c('x1', 'x2'), data=dat2, q=2, family=binomial()))
(qfit2ee <- qgcomp.emm.glm.ee(f=y ~ z + x1 + x2, emmvar="z",
  expnms = c('x1', 'x2'), data=dat2, q=2, family=binomial(), rr=FALSE))
# Note under rr = TRUE that the risk ratio will be reported in the MSM results
(qfit2eerr <- qgcomp.emm.glm.ee(f=y ~ z + x1 + x2, emmvar="z",
  expnms = c('x1', 'x2'), data=dat2, q=2, family=binomial(), rr=TRUE))
# linear model, factor modifier
dat <- data.frame(y=runif(150), x1=runif(150), x2=runif(150),
  r=rbinom(150, 1, 0.5), z=sample(c(1, 2, 3), 150, replace=TRUE))
#note need to declare factor
dat$zfact = as.factor(dat$z)
# this can fail if the model is unidentified (e.g. z and zfact are included in the model)
(qfit <- qgcomp.emm.glm.noboot(f=y ~ zfact + x1 + x2, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
(qfitee <- qgcomp.emm.glm.ee(f=y ~ zfact + x1 + x2, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
library(qgcomp)
# standard qgcomp model without interaction
(qfitee_noemm <- qgcomp.glm.ee(f=y ~ zfact + x1 + x2,
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
  qfitee_noemm$fit # underlying fit
# global test for interaction
anova(qfitee, qfitee_noemm)
# get stratified effect estimates:
getstrateffects(qfitee, emmval=1)
getstrateffects(qfitee, emmval=2)
getstrateffects(qfitee, emmval=3)
dat$rfact = as.factor(dat$r)
(qfiteer <- qgcomp.emm.glm.ee(f=y ~ zfact + x1 + x2 + r, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
 # factor as a confounder, also works if the modifier is not in the model
(qfiteerr2 <- qgcomp.emm.glm.ee(f=y ~  x1 + x2 + rfact, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=2, family=gaussian()))
getstrateffects(qfiteerr2, emmval=2)
getjointeffects(qfiteerr2, emmval=2)
modelbound(qfiteerr2, emmval=2)
pointwisebound(qfiteerr2, emmval=2)
(qfitee <- qgcomp.glm.ee(f=y ~ zfact + x1 + x2, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=10, degree=2, family=gaussian()))
(qfitee <- qgcomp.emm.glm.ee(f=y ~ zfact + x1 + x2, emmvar="zfact",
  expnms = c('x1', 'x2'), data=dat, q=10, degree=2, family=gaussian()))