impute.R

baseline = read.csv('baseline_data.csv')
origpredictors = read.csv('baseline_predictors.csv')$predictors
predictors = read.csv('baseline_predictors.csv')$predictors
#alter predictors to include RaceEthnicity and not the levels
predictors = predictors[!(predictors %in% c("Asian","Black","Hispanic","Other"))]
predictors = c(predictors,"RaceEthnicity")

sleep = c("sds_p_ss_dims","sds_p_ss_sbd","sds_p_ss_da","sds_p_ss_swtd","sds_p_ss_does","sds_p_ss_shy")

summary(baseline[,predictors])
summary(baseline[,sleep])

mean(100*colSums(is.na(baseline[,predictors]))/nrow(baseline))

i = rowSums(is.na(baseline[,predictors]))
j = rowSums(is.na(baseline[,sleep]))

drop = j==6

dat = baseline[!drop,c("subjectkey","eventname",sleep,predictors)]
#correctly code categorical, etc
dat$demo_comb_income_v2 = factor(dat$demo_comb_income_v2,levels=c(1:10))
dat$M = factor(dat$M,levels=c(0,1))
dat$RaceEthnicity = factor(dat$RaceEthnicity,levels=c("White","Asian","Black","Hispanic","Other"))
dat$not_married = factor(dat$not_married,levels=c(0,1))
dat$medhx_2a = factor(dat$medhx_2a,levels=c(0,1))
dat$medhx_2d = factor(dat$medhx_2d,levels=c(0,1))
dat$medhx_ss_4b_p = factor(dat$medhx_ss_4b_p,levels=c(0,1,2,3,4,5))
dat$medhx_ss_5b_p = factor(dat$medhx_ss_5b_p,levels=c(0,1,2,3,4,5))
dat$neighborhood1r_p = factor(dat$neighborhood1r_p,levels=c(1,2,3,4,5))
dat$neighborhood2r_p = factor(dat$neighborhood2r_p,levels=c(1,2,3,4,5))
dat$neighborhood3r_p = factor(dat$neighborhood3r_p,levels=c(1,2,3,4,5))

mat = dat[,c(sleep, predictors)]

library(mice)

blocks = list(sleep=sleep, predictors=predictors)
predictorMatrix = matrix(c(rep(1,length(sleep)),rep(0,length(predictors)),rep(0,length(sleep)),rep(1,length(predictors))),nrow=2,byrow=T)
seed = 1234

m = mice(mat,m=50,predictorMatrix = predictorMatrix, blocks = blocks, seed= seed, maxit=20)
plot(m)

ymean = matrix(NA,nrow=length(sleep),ncol=50)
xmean = matrix(NA,nrow=length(origpredictors),ncol=50)

for (i in 1:50) {
  tmp = complete(m,i)
  tmp$subjectkey = dat$subjectkey
  write.csv(tmp,paste0("imputed_",i,".csv"),row.names=F)
}