MImix (version 1.0)

diaph.data: Diaphanography Partial Verification Bias Example

Description

This dataset consists of a list of imputed datasets for an example of multiple imputation for the correction of partial verification bias, as originally reported in Harel and Zhou (2006) and discussed by deGroot, et al. (2008).

Usage

diaph.data

Format

A list containing three elements: imputed.tables: A list of 100 imputed tables using the saturated categorical model of Schafer (1997). sens.samples: A vector of 1000 draws of the sensitivity of the test from the posterior distribution using the saturated categorical model of Schafer (1997). original.data: The original dataset from Marshall, et al. (1981), including the observations with missing gold standard values.

Source

Marshall, V., Williams, D.C., and Smith, K. D. (1981). Diaphanography as a means of detecting breast cancer. Radiology 150:339-343.

Details

For the imputed tables and the original table, D refers to the gold standard test and T refers to the diagnostic test of interest.

References

Schafer, J. L. (1997). Analysis of Incomplete Multivariate Data by Simulation. Chapman & Hall Ltd.

Harel O., Zhou X.H. (2006) Multiple imputation for correcting verification bias. Statistics in Medicine 25:3769-3786.

de Groot, J.A.H. and Janssen, K.J.M. and Zwinderman, A.H. and Moons, K.G.M. and Reitsma, J.B. (2008) Multiple imputation to correct for partial verification bias revisited. Statistics in Medicine 27:5880-5899.

Examples

data(diaph.data)
attach(diaph.data)
## Calculate sensitivity for each imputed table

sens.imps<-lapply(imputed.tables,function(x){ x[2,2]/(x[2,1]+x[2,2]) })
sens.imps.vars<-lapply(imputed.tables,function(x){ x[2,2]*x[2,1]/(x[2,1]+x[2,2])^3 })

### Calculate mixture summary

MImix(sens.imps,sens.imps.vars)

### Compare to usual t-summary using MIcombine: requires(mitools)

library(mitools)
summary(MIcombine(sens.imps,sens.imps.vars))

### Compare both to the Bayesian posterior estimate

quantile(sens.samples,c(0.025,0.5,0.975))