# Use the 2000 EDHS Dataset
library(haven)
library(tidyr)
library(dplyr)
library(lme4)
library(nlme)


# 1st read the csv file, BMI_obov is the outcome var. 
ETIR41FL <- read.csv("ETIR41FL_ovob.csv")

# Filtering the non-flagged casses of BMI(v445), the non-pregnant women(v213) 
# and non-postpartum women, (v222) from the whole dataset. 
# ETIR41FL_M <- ETIR41FL %>% filter(v445!=9998, v213=="0", v222>"2")
# View(ETIR41FL_M)

#Removing all the NA(take only the completed cases)
ETIR41FL_N <- ETIR41FL_M[which(complete.cases
                               (ETIR41FL_M[,c('AgeCat','v106','v130','v313','MaritalSt','v714','Freq_newsp',
                                              'Freq_radio','Freq_tv','Parity','Partner_edu',
                                              'v024','v025','Com_edu')])),]

#2nd Model 
Nutri_Multi2 <- glmer(BMI_obov2~as.factor(AgeCat)+as.factor(v106)+as.factor(v130)+
                        as.factor(v313)+ as.factor(MaritalSt)+ as.factor(v714)+ 
                        as.factor(Freq_newsp)+  as.factor(Freq_radio)+  
                        as.factor(Freq_tv)+ as.factor(Parity)+ 
                        as.factor(Partner_edu)+ (1|v024), family=binomial, 
                      weights=ETIR41FL_N$WGT, data=ETIR41FL_N, 
                      na.action=na.omit)
summary(Nutri_Multi2)

# 3rd Model II (containing only community level factors) and we changed the previous community
# level variable Com_urbanP by Com_PlRes.

Nutri_Multi3 <- glmer(BMI_obov2~as.factor(v024)+as.factor(v025)+ as.factor(Com_edu)+ 
                      (1|v024), family=binomial,weights=ETIR41FL_N$WGT, 
                      data=ETIR41FL_N, na.action=na.omit)
summary(Nutri_Multi3)