library(mice) ## for md.pattern and for data set "boys"
## data
?boys
str(boys)
head(boys)

##################################################
## Look at missing values
##################################################
md.pattern(boys) ## 223 rows are complete
223/748 ## only 30% of rows are complete
1622/(748*9) ## 24% missing values in total

## Just for illustrative purposes: 
## We want to explain the head circumference (hc) by age, hgt and wgt

##################################################
## Default of lm if missing values are present: na.omit
##################################################
fmDefault <- lm(hc ~ age + hgt + wgt, data = boys)
summary(fmDefault)
?lm
?na.omit
## is the same as:
fmDefault <- lm(hc ~ age + hgt + wgt, data = boys, na.action = na.omit)
summary(fmDefault)
## set to other default
fmDefault <- lm(hc ~ age + hgt + wgt, data = boys, na.action = na.fail)

##################################################
## complete case analysis by hand
##################################################
dfBoys <- boys[,c(1,2,3,5)] ## select variables: age, hgt, wgt, hc
?complete.cases
complete.cases(dfBoys)
datCC <- boys[complete.cases(dfBoys),]
dim(boys)
dim(datCC)
## gives the same result as with na.omit
fmCC <- lm(hc ~ age + hgt + wgt, data = datCC)
summary(fmCC)

##################################################
## Single Imputation (Conditional Mean imputation using missForest)
##################################################
str(boys)
library(missForest)
set.seed(32)
impMF <- missForest(boys)
impMF$OOBerror
datImp <- impMF$ximp
str(datImp)
fmMF <- lm(hc ~ age + hgt + wgt, data = datImp)
summary(fmMF)