qgcomp (version 2.2.2)

mice.impute.leftcenslognorm: imputation for limits of detection problems

Description

This function integrates with mice to impute values below the LOD using a left censored log-normal distribution.

Usage

mice.impute.leftcenslognorm(
  y,
  ry,
  x,
  wy = NULL,
  lod = NULL,
  debug = FALSE,
  ...
)

Arguments

y

Vector to be imputed

ry

Logical vector of length length(y) indicating the the subset y[ry] of elements in y to which the imputation model is fitted. The ry generally distinguishes the observed (TRUE) and missing values (FALSE) in y.

x

Numeric design matrix with length(y) rows with predictors for y. Matrix x may have no missing values.

wy

Logical vector of length length(y). A TRUE value indicates locations in y for which imputations are created.

lod

numeric vector of limits of detection (must correspond to index in original data)

debug

logical, print extra info

...

arguments to survreg

Value

Vector with imputed data, same type as y, and of length sum(wy)

Details

While this function has utility far beyond qgcomp, it is included in the qgcomp package because it will be useful for a variety of settings in which qgcomp is useful. Note that LOD problems where the LOD is small, and the q parameter from qgcomp.noboot or qgcomp.boot is not large, the LOD may be below the lowest quantile cutpoint which will yield identical datasets from the MICE procedure in terms of quantized exposure data. If only exposures are missing, and they have low LODs, then there will be no benefit in qgcomp from using MICE rather than imputing some small value below the LOD.

This function will likely be relegated to a separate package in the future

Examples

# NOT RUN {
N = 100
set.seed(123)
dat <- data.frame(y=runif(N), x1=runif(N), x2=runif(N), z=runif(N))
true = qgcomp.noboot(f=y ~ z + x1 + x2, expnms = c('x1', 'x2'), 
        data=dat, q=2, family=gaussian())
mdat <- dat
mdat$x1 = ifelse(mdat$x1>0.5, mdat$x1, NA)
mdat$x2 = ifelse(mdat$x2>0.75, mdat$x2, NA)
cc <- qgcomp.noboot(f=y ~ z + x1 + x2, expnms = c('x1', 'x2'), 
       data=mdat[complete.cases(mdat),], q=2, family=gaussian())

# }
# NOT RUN {
# note the following example imputes from the wrong parametric model and is expected to be biased
# as a result (but it demonstrates how to use qgcomp and mice together)
library("mice")
library("survival")
impdat = mice(data = mdat, 
  method = c("", "leftcenslognorm", "leftcenslognorm", ""),
  lod=c(NA, 0.5, 0.75, NA), debug=FALSE)
qc.fit.imp <- list(
  call = call("qgcomp.noboot(y~., expnms = c('x1', 'x2'), family=gaussian())"),
  call1 = impdat$call,
  nmis = impdat$nmis,
  analyses = lapply(1:5, function(x) qgcomp.noboot(y~., expnms = c("x1", "x2"),
    data=complete(impdat, x), family=gaussian(), bayes=TRUE))
)
obj = pool(as.mira(qc.fit.imp))
# true values
true
# complete case analysis
cc
# MI based analysis
summary(obj)
# }