qgcomp.zi.boot  R Documentation 
This function estimates a linear doseresponse parameter representing a one quantile increase in a set of exposures of interest for zeroinflated count outcomes. This function is limited to linear and additive effects of individual components of the exposure. This model estimates the parameters of a marginal structural zeroinflated count model (MSM) based on gcomputation with quantized exposures. Note: this function allows linear and nonadditive effects of individual components of the exposure, as well as nonlinear joint effects of the mixture via polynomial basis functions, which increase the computational computational burden due to the need for nonparametric bootstrapping.
qgcomp.zi.boot(
f,
data,
expnms = NULL,
q = 4,
breaks = NULL,
id = NULL,
weights,
alpha = 0.05,
B = 200,
degree = 1,
seed = NULL,
bayes = FALSE,
parallel = FALSE,
MCsize = 10000,
msmcontrol = zimsm_fit.control(),
parplan = FALSE,
...
)
f 
R style formula 
data 
data frame 
expnms 
character vector of exposures of interest 
q 
NULL or number of quantiles used to create quantile indicator variables representing the exposure variables. If NULL, then gcomp proceeds with untransformed version of exposures in the input datasets (useful if data are already transformed, or for performing standard gcomputation) 
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) 
weights 
"case weights"  passed to the "weight" argument of

alpha 
alpha level for confidence limit calculation 
B 
integer: number of bootstrap iterations (this should typically be >=200, though it is set lower in examples to improve runtime). 
degree 
polynomial basis function for marginal model (e.g. degree = 2 allows that the relationship between the whole exposure mixture and the outcome is quadratic.) 
seed 
integer or NULL: random number seed for replicable bootstrap results 
bayes 
not currently implemented. 
parallel 
use (safe) parallel processing from the future and future.apply packages 
MCsize 
integer: sample size for simulation to approximate marginal zero inflated model parameters. This can be left small for testing, but should be as large as needed to reduce simulation error to an acceptable magnitude (can compare psi coefficients for linear fits with qgcomp.zi.noboot to gain some intuition for the level of expected simulation error at a given value of MCsize) 
msmcontrol 
named list from 
parplan 
(logical, default=FALSE) automatically set future::plan to plan(multisession) (and set to existing plan, if any, after bootstrapping) 
... 
arguments to glm (e.g. family) 
Zeroinflated count models allow excess zeros in standard count outcome (e.g. Poisson distributed outcomes). Such models have two components: 1 ) the probability of arising from a degenerate distribution at zero (versus arising from a count distribution) and 2 ) the rate parameter of a count distribution. Thus, one has the option of allowing exposure and covariate effects on the zero distribution, the count distribution, or both. The zero distribution parameters correspond to logodds ratios for the probability of arising from the zero distribution. Count distribution parameters correspond to lograteratio parameters. Test statistics and confidence intervals are based on a nonparametric bootstrap, using the standard deviation of the bootstrap estimates to estimate the standard error. The bootstrap standard error is then used to estimate Waldtype confidence intervals. Note that no bootstrapping is done on estimated quantiles of exposure, so these are treated as fixed quantities.
Of note, this function yields marginal estimates of the expected outcome under values of the joint exposure quantiles (e.g. the expected outcome if all exposures are below the 1st quartile). These outcomes can be used to derive estimates of the effect on the marginal expectation of the outcome, irrespective of zeroinflated/count portions of the statistical model.
Estimates correspond to the average expected change in the (log) outcome per quantile increase in the joint exposure to all exposures in ‘expnms’. Test statistics and confidence intervals are based on a nonparametric bootstrap, using the standard deviation of the bootstrap estimates to estimate the standard error. The bootstrap standard error is then used to estimate Waldtype confidence intervals. Note that no bootstrapping is done on estimated quantiles of exposure, so these are treated as fixed quantities
a qgcompfit object, which contains information about the effect measure of interest (psi) and associated variance (var.psi), as well as information on the model fit (fit) and information on the marginal structural model (msmfit) used to estimate the final effect estimates.
Other qgcomp_methods:
qgcomp.cch.noboot()
,
qgcomp.cox.boot()
,
qgcomp.cox.noboot()
,
qgcomp.glm.boot()
,
qgcomp.glm.noboot()
,
qgcomp.hurdle.boot()
,
qgcomp.hurdle.noboot()
,
qgcomp.multinomial.boot()
,
qgcomp.multinomial.noboot()
,
qgcomp.partials()
,
qgcomp.zi.noboot()
set.seed(50)
n=100
dat < data.frame(y=rbinom(n, 1, 0.5)*rpois(n, 1.2), x1=runif(n), x2=runif(n), z=runif(n))
# poisson count model, mixture in both portions
## Not run:
# warning: the examples below can take a long time to run
res = qgcomp.zi.boot(f=y ~ x1 + x2  x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="poisson", B=1000, MCsize=10000, parallel=TRUE, parplan=TRUE)
qgcomp.zi.noboot(f=y ~ x1 + x2  x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="poisson")
res
# accuracy for small MCsize is suspect (compare coefficients between boot/noboot versions),
# so recheck with MCsize set to larger value (this takes a long time to run)
res2 = qgcomp.zi.boot(f=y ~ x1 + x2  x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="poisson", B=1000, MCsize=50000, parallel=TRUE, parplan=TRUE)
res2
plot(density(res2$bootsamps[4,]))
# negative binomial count model, mixture and covariate in both portions
qgcomp.zi.boot(f=y ~ z + x1 + x2  z + x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="negbin", B=10, MCsize=10000)
# weighted analysis (NOTE THIS DOES NOT WORK WITH parallel=TRUE!)
dat$w = runif(n)*5
qgcomp.zi.noboot(f=y ~ z + x1 + x2  x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="poisson", weights=w)
# Expect this:
# Warning message:
# In eval(family$initialize) : noninteger #successes in a binomial glm!
qgcomp.zi.boot(f=y ~ x1 + x2  x1 + x2, expnms = c('x1', 'x2'),
data=dat, q=4, dist="poisson", B=5, MCsize=50000, parallel=FALSE, weights=w)
# Log rr per one IQR change in all exposures (not on quantile basis)
dat$x1iqr < dat$x1/with(dat, diff(quantile(x1, c(.25, .75))))
dat$x2iqr < dat$x2/with(dat, diff(quantile(x2, c(.25, .75))))
# note that I(x>...) now operates on the untransformed value of x,
# rather than the quantized value
res2 = qgcomp.zi.boot(y ~ z + x1iqr + x2iqr + I(x2iqr>0.1) + I(x2>0.4) + I(x2>0.9)  x1iqr + x2iqr,
family="binomial", expnms = c('x1iqr', 'x2iqr'), data=dat, q=NULL, B=2,
degree=2, MCsize=200, dist="poisson")
res2
## End(Not run)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.