With themiceafter package you can apply statistical and pooled analyses after multiple imputation. Therefore the name ‘miceafter’. The package contains a variety of statistical tests like thepool_levenetest function to pool Levene’s tests across multiply imputed datasets or thepool_propdiff_nw function to pool the difference between proportions according to method Newcombe-Wilson. The package also contains a functionpool_glm to pool and select linear and logistic regression functions. Functions can also be used in combination with the%>% (Pipe) operator.
More and more statistical analyses and pooling functions will be added over time to form a framework of statistical tests that can be applied and pooled across multiply imputed datasets.
Pooling Levene’s Test
This example shows you how to pool the Levene test across 5 multiply imputed datasets. The pooling method that is used is method D1.
library(miceafter)# Step 1: Turn data frame with multiply imputed datasets into object of 'milist'imp_dat<-df2milist(lbpmilr, impvar="Impnr")# Step 2: Do repeated analyses across multiply imputed datasetsra<-with(imp_dat, expr=levene_test(Pain~factor(Carrying)))# Step 3: Pool repeated test resultsres<-pool_levenetest(ra, method="D1")res#> F_value df1 df2 P(>F) RIV#> [1,] 1.586703 2 115.3418 0.209032 0.1809493#> attr(,"class")#> [1] "mipool"Pooling Difference between Proportions according to Newcombe-Wilson
library(miceafter)# Step 1: Turn data frame with multiply imputed datasets into object of 'milist'imp_dat<-df2milist(lbpmilr, impvar="Impnr")# Step 2: Do repeated analyses across multiply imputed datasetsra<-with(imp_dat, expr=propdiff_wald(Chronic~Radiation, strata=TRUE))# Step 3: Pool repeated test resultsres<-pool_propdiff_nw(ra)res#> Prop diff CI L NW CI U NW#> [1,] 0.2786 0.1199 0.419#> attr(,"class")#> [1] "mipool"See for more functions thepackage website
Main functions
The main functions of the package are thedf2milist,list2milist,mids2milist and thewith.milist functions. Thedf2milist function turns a data frame with multiply imputed datasets into an object of classmilist, thelist2milist does this for a list with multiply imputed datasets and themids2milist for objects of classmids. Thesemilist object can than be used with thewith.milist function to apply repeated statistical analyses across the multiply imputed datasets. Subsequently, pooling functions are available in the form of separatepool functions.
Installation
You can install the development version fromGitHub with:
# install.packages("devtools")devtools::install_github("mwheymans/miceafter")License
- GPL (>= 2)
Citation
Developers
- Martijn Heymans
Maintainer, author - More about authors...
