library(micsr)

micsr.data::afghan_girls

afghan <- read.csv("afghan.csv")
names(afghan)
with(afghan, table(treat))

round(aggregate(cbind(age_child, age_head_child, edu_head,
                      land, animals, Time, n_people, near_school) ~ treat, 
                data = afghan, FUN = mean), 2)

round(c(t.test(age_child ~ treat, data = afghan, var.equal = TRUE)$p.value,
        t.test(age_head_child ~ treat, data = afghan, var.equal =TRUE)$p.value,
        t.test(edu_head ~ treat, data = afghan, var.equal = TRUE)$p.value,
        t.test(land ~ treat, data = afghan, var.equal = TRUE)$p.value,
        t.test(animals ~ treat, data = afghan, var.equal = TRUE)$p.value,
        t.test(Time~ treat, data = afghan, var.equal = TRUE)$p.value,          
        t.test(n_people ~ treat, data = afghan, var.equal = TRUE)$p.value,
        t.test(near_school ~ treat, data = afghan, var.equal =TRUE)$p.value),2)

round(c(t.test(age_child ~ treat, data = afghan)$p.value,
        t.test(age_head_child ~ treat, data = afghan)$p.value,
        t.test(edu_head ~ treat, data = afghan)$p.value,
        t.test(land ~ treat, data = afghan)$p.value,
        t.test(animals ~ treat, data = afghan)$p.value,
        t.test(Time~ treat, data = afghan)$p.value,          
        t.test(n_people ~ treat, data = afghan)$p.value,
        t.test(near_school ~ treat, data = afghan)$p.value),2)


t.test(near_school ~ treat, data = afghan, var.equal =TRUE)

summary(lm(near_school ~ treat, afghan))$coefficients 


n <- as.numeric(with(afghan, table(treat)))
round(rbind(with(afghan, table(heads_child, treat))[2,],
            with(afghan, table(heads_child, treat))[2,]/n),2)

round(rbind(with(afghan, table(girls, treat))[1,],
            with(afghan, table(girls, treat))[1,]/n),2)

round(rbind(with(afghan, table(girls, treat))[2,],
            with(afghan, table(girls, treat))[2,]/n),2)


round(rbind(with(afghan, table(farmer, treat))[1,],
            with(afghan, table(farmer, treat))[1,]/n),2)

round(rbind(with(afghan, table(farmer, treat))[2,],
            with(afghan, table(farmer, treat))[2,]/n),2)


afghan$ethny <- "Other"
afghan$ethny[afghan$farsi == 1] <- "farsi" 
afghan$ethny[afghan$tajik == 1] <- "tajik" 
round(rbind(with(afghan, table(ethny, treat))[1,],
            with(afghan, table(ethny, treat))[1,]/n),2)

round(rbind(with(afghan, table(ethny, treat))[2,],
            with(afghan, table(ethny, treat))[2,]/n),2)

round(rbind(with(afghan, table(ethny, treat))[3,],
            with(afghan, table(ethny, treat))[3,]/n),2)


obs_freq <- prop.table(table(afghan$treat, afghan$heads_child))

freq_treat <- prop.table(table(afghan$treat))
freq_heads_child <- prop.table(table(afghan$heads_child))
ind_freq <- outer(freq_treat, freq_heads_child)

sum((obs_freq - ind_freq) ^ 2 / ind_freq) * nrow(afghan)

chisq.test(table(afghan$treat, afghan$heads_child), correct = FALSE) 

pchisq(2.970317, 1, lower.tail = FALSE)


obs_freq <- prop.table(table(afghan$treat, afghan$girls))

freq_treat <- prop.table(table(afghan$treat))
freq_girls <- prop.table(table(afghan$girls))
ind_freq <- outer(freq_treat, freq_girls)

sum((obs_freq - ind_freq) ^ 2 / ind_freq) * nrow(afghan)


chisq.test(table(afghan$treat, afghan$girls), correct = FALSE) 

pchisq(0.5750879, 1, lower.tail = FALSE)


obs_freq <- prop.table(table(afghan$treat, afghan$farmer))

freq_treat <- prop.table(table(afghan$treat))
freq_farmer <- prop.table(table(afghan$farmer))
ind_freq <- outer(freq_treat, freq_farmer)

sum((obs_freq - ind_freq) ^ 2 / ind_freq) * nrow(afghan)

chisq.test(table(afghan$treat, afghan$farmer), correct = FALSE) 

pchisq(0.1855524, 1, lower.tail = FALSE)


obs_freq <- prop.table(table(afghan$treat, afghan$ethny))

freq_treat <- prop.table(table(afghan$treat))
freq_ethny <- prop.table(table(afghan$ethny))
ind_freq <- outer(freq_treat, freq_ethny)

sum((obs_freq - ind_freq) ^ 2 / ind_freq) * nrow(afghan)

chisq.test(table(afghan$treat, afghan$ethny), correct = FALSE) 

pchisq(2.846, 2, lower.tail = FALSE)

par(mfrow=c(1,2))
with(afghan[afghan$girls==1,], boxplot(score ~ treat))
with(afghan[afghan$girls == 0, ], boxplot(score ~ treat))

t.test(score ~ treat, data = afghan, subset = (girls == 1), 
       var.equal = TRUE)

mean(afghan$score[afghan$treat == 1 & afghan$girls == 1], na.rm = TRUE) - mean(afghan$score[afghan$treat == 0 & afghan$girls == 1], , na.rm = TRUE)


fit <- lm(score ~ treat, data = afghan, subset = (girls==1))
summary(fit)


  
fit2 <- lm(score ~ treat + Time + n_people + animals + near_school + 
             heads_child + age_child + age_head_child + 
             farmer + land + ethny + chagcharan + edu_head, data = afghan, subset = (girls==1))
summary(fit2)