library(ordinal)

ratings = read.csv('Ratings_Exp2.csv',header=TRUE)

# Remove dyslexic subjects and subjects who did not respond to at least two items in each condition
to_remove = c(7,13,38,42,87)  # dyslexic
to_remove = c(to_remove,1,2,5:9,11,12,15,17:19,28:74,76,78,79,82:84,97,98,100:106,108,112,117:119,121:126)   # did not respond to enough items

for (s in to_remove) {
  ratings = ratings[ratings$Subject != s,]
}

# Get condition means and by-subject SEs
means = data.frame(Aspect = c("Comprehensibility","Acceptability"),
                   V0 = c(mean(ratings[ratings$Condition=="0",]$Comprehensibility,na.rm=TRUE),mean(ratings[ratings$Condition=="0",]$Acceptability,na.rm=TRUE)),
                   V1 = c(mean(ratings[ratings$Condition=="1",]$Comprehensibility,na.rm=TRUE),mean(ratings[ratings$Condition=="1",]$Acceptability,na.rm=TRUE)),
                   V2 = c(mean(ratings[ratings$Condition=="2",]$Comprehensibility,na.rm=TRUE),mean(ratings[ratings$Condition=="2",]$Acceptability,na.rm=TRUE)),
                   V3 = c(mean(ratings[ratings$Condition=="3",]$Comprehensibility,na.rm=TRUE),mean(ratings[ratings$Condition=="3",]$Acceptability,na.rm=TRUE)))

subj_id   = unique(ratings$Subject)
nr_subj   = length(subj_id)
subj_mean = matrix(nrow=8,ncol=nr_subj)
p = 1
for (s in subj_id) {
  subj_mean[1,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="0",]$Comprehensibility,na.rm=TRUE)
  subj_mean[2,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="1",]$Comprehensibility,na.rm=TRUE)
  subj_mean[3,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="2",]$Comprehensibility,na.rm=TRUE)
  subj_mean[4,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="3",]$Comprehensibility,na.rm=TRUE)
  subj_mean[5,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="0",]$Acceptability,na.rm=TRUE)
  subj_mean[6,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="1",]$Acceptability,na.rm=TRUE)
  subj_mean[7,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="2",]$Acceptability,na.rm=TRUE)
  subj_mean[8,p] = mean(ratings[ratings$Subject==s & ratings$Condition=="3",]$Acceptability,na.rm=TRUE)
  p = p+1
}

SEs = data.frame(Aspect = c("Comprehensibility","Acceptability"),
                 V0 = c(sd(subj_mean[1,])/sqrt(nr_subj),sd(subj_mean[5,])/sqrt(nr_subj)),
                 V1 = c(sd(subj_mean[2,])/sqrt(nr_subj),sd(subj_mean[6,])/sqrt(nr_subj)),
                 V2 = c(sd(subj_mean[3,])/sqrt(nr_subj),sd(subj_mean[7,])/sqrt(nr_subj)),
                 V3 = c(sd(subj_mean[4,])/sqrt(nr_subj),sd(subj_mean[8,])/sqrt(nr_subj)))

# Turn subject, item and condition numbers into factors
ratings$Subject   = as.factor(ratings$Subject)
ratings$Item      = as.factor(ratings$Item)
ratings$Condition = as.factor(ratings$Condition)

# Turn scores into ordinal variable
ratings$Acceptability     = ordered(ratings$Acceptability)
ratings$Comprehensibility = ordered(ratings$Comprehensibility)

# Run ordinal mixed-effects analyses
model_acceptability     = clmm(Acceptability     ~ Condition  + (1+Condition|Subject) + (1+Condition|Item), data=ratings)
model_comprehensibility = clmm(Comprehensibility ~ Condition  + (1+Condition|Subject) + (1+Condition|Item), data=ratings)

rm(ratings,subj_id,subj_mean,nr_subj,p,s)
save.image('Exp2.RData')