Working together: mice and psfmi

Martijn W Heymans

2021-01-13

Introduction

The mice function is one of the most used functions to apply multiple imputation. This page shows how functions in the psfmi package can be easily used in combination with mice. In this way multivariable models can easily be developed in combination with mice.

Installing the psfmi and mice packages

You can install the released version of psfmi with:

install.packages("psfmi")

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("mwheymans/psfmi")

You can install the released version of mice with:

install.packages("mice")

Examples

mice and psfmi for pooling logistic regression models


  library(psfmi)
  library(mice)

  imp <- mice(lbp_orig, m=5, maxit=5) 
#> 
#>  iter imp variable
#>   1   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
  
  data_comp <- complete(imp, action = "long", include = FALSE)
  
  library(psfmi)
  pool_lr <- psfmi_lr(data=data_comp, nimp=5, impvar=".imp", Outcome="Chronic",
  predictors=c("Gender", "Smoking", "Function", "JobControl",
  "JobDemands", "SocialSupport"), method="D1")
  
  pool_lr$RR_model
#> $`Step 1 - no variables removed -`
#>            term     estimate  std.error  statistic       df     p.value
#> 1   (Intercept)  0.256763455 2.43312414  0.1055283 134.4059 0.916113945
#> 2        Gender -0.359441075 0.41876991 -0.8583259 142.0828 0.392158157
#> 3       Smoking  0.067952121 0.33814663  0.2009546 149.5614 0.841007100
#> 4      Function -0.138169108 0.04320364 -3.1980896 138.9868 0.001713409
#> 5    JobControl  0.007638501 0.01972951  0.3871612 137.4752 0.699235587
#> 6    JobDemands -0.005389408 0.03704788 -0.1454714 137.6611 0.884551476
#> 7 SocialSupport  0.040056018 0.05692914  0.7036119 137.4538 0.482864689
#>          OR lower.EXP   upper.EXP
#> 1 1.2927393 0.0105103 159.0035655
#> 2 0.6980664 0.3050542   1.5974103
#> 3 1.0703141 0.5486967   2.0878061
#> 4 0.8709514 0.7996426   0.9486193
#> 5 1.0076677 0.9691130   1.0477563
#> 6 0.9946251 0.9243672   1.0702231
#> 7 1.0408691 0.9300526   1.1648894

Back to Examples

mice and psfmi for selecting logistic regression models


  library(psfmi)
  library(mice)

  imp <- mice(lbp_orig, m=5, maxit=5) 
#> 
#>  iter imp variable
#>   1   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   1   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   2   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   3   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   4   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   1  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   2  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   3  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   4  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
#>   5   5  Carrying  Pain  Tampascale  Function  Radiation  Age  Satisfaction  JobControl  JobDemands  SocialSupport
  
  data_comp <- complete(imp, action = "long", include = FALSE)
  
  library(psfmi)
  pool_lr <- psfmi_lr(data=data_comp, nimp=5, impvar=".imp", Outcome="Chronic",
  predictors=c("Gender", "Smoking", "Function", "JobControl",
  "JobDemands", "SocialSupport"), p.crit = 0.157, method="D1",
  direction = "FW")
#> Entered at Step 1 is - Function
#> 
#> Selection correctly terminated, 
#> No new variables entered the model
  
  pool_lr$RR_model_final
#> $`Final model`
#>          term   estimate  std.error statistic       df     p.value        OR
#> 1 (Intercept)  1.2331799 0.47595922  2.590936 115.4096 0.010806296 3.4321260
#> 2    Function -0.1399743 0.04239566 -3.301620 107.6885 0.001304682 0.8693805
#>   lower.EXP upper.EXP
#> 1  1.337005 8.8103567
#> 2  0.799305 0.9455996

Back to Examples