Multiple imputation of missing data with mice

Example by Jerry Reiter for short course at Odum Institute, March 2018. This short example shows R commands for handling missing data using the mice package.

Preliminaries (load packages and data)

If you have not already, install the mice package with the following command (remove #):

#install.packages("mice")

Once you have mice installed, load it into memory using the command:

library(mice)

## Loading required package: lattice

## 
## Attaching package: 'mice'

## The following objects are masked from 'package:base':
## 
##     cbind, rbind

We illustrate the package with a simple data set from the mice package:

data(nhanes2)

dim(nhanes2)

## [1] 25  4

summary(nhanes2)

##     age          bmi          hyp          chl       
##  20-39:12   Min.   :20.40   no  :13   Min.   :113.0  
##  40-59: 7   1st Qu.:22.65   yes : 4   1st Qu.:185.0  
##  60-99: 6   Median :26.75   NA's: 8   Median :187.0  
##             Mean   :26.56             Mean   :191.4  
##             3rd Qu.:28.93             3rd Qu.:212.0  
##             Max.   :35.30             Max.   :284.0  
##             NA's   :9                 NA's   :10

Note that the age and hyp variables are already defined as factor variables in this dataset. If you have a dataset with categorical variables without this pre-definition, you should make the variables factors before doing mice. For example, to make a variable called var1 into a factor in some dataset called thedata, type thedata$var1 = as.factor(thedata$var1).

Missing data

The NA values represent missing data.

Let’s look at the missing data pattern:

md.pattern(nhanes2)

##    age hyp bmi chl   
## 13   1   1   1   1  0
## 3    1   1   1   0  1
## 1    1   1   0   1  1
## 1    1   0   0   1  2
## 7    1   0   0   0  3
##      0   8   9  10 27

Exploratory data analyses

Let’s look at some exploratory data analyses based on the complete cases.

First make the complete cases

ccnhanes2data = cc(nhanes2)
ccnhanes2data

##      age  bmi hyp chl
## 2  40-59 22.7  no 187
## 5  20-39 20.4  no 113
## 7  20-39 22.5  no 118
## 8  20-39 30.1  no 187
## 9  40-59 22.0  no 238
## 13 60-99 21.7  no 206
## 14 40-59 28.7 yes 204
## 17 60-99 27.2 yes 284
## 18 40-59 26.3 yes 199
## 19 20-39 35.3  no 218
## 22 20-39 33.2  no 229
## 23 20-39 27.5  no 131
## 25 40-59 27.4  no 186

Note that there are only 3 cases with hypertension = yes. Relationships will be hard to model accurately due to small sample size…

We will use these data for illustration of the MI procedures. Really we need more data to say anything meaningful about relationships among these variables.

plot(ccnhanes2data$bmi, x=ccnhanes2data$chl, xlab = "Cholesterol", ylab = "BMI", main = "Complete cases: bmi versus chl")

boxplot(ccnhanes2data$bmi ~ ccnhanes2data$age, xlab = "Age", ylab = "BMI", main = "Complete cases: bmi versus age")

boxplot(ccnhanes2data$bmi ~ ccnhanes2data$hyp, xlab = "Hypertensive", ylab = "BMI", main = "Complete cases: bmi versus hyp")