# input format
# 1) subject id,
# 2) sentence id,
# 3) CONDITION,
# 4) word number (the word which was fixated), 2=PoF, 3=n+1, 4=n+2,
# 5) LAND POS OF 1ST
# 6) GAZE duration
# 7) NUM OF FIX
# 8) First Fix Dur
# 9) out going saccade length
# 10) LAND POS OF LAST
# 11) incoming saccade length

rm(list=ls())
library(lme4)
load('chs_np2_09_new.rda')

# center
cen <- c(16:21)
gd[ ,cen]  <- scale(gd[ ,cen], scale=F)

gd$sn <- factor(gd$sn)
gd$id <- factor(gd$id)

# Dur filter
gd.ffd = gd[which(gd[,8]>60 & gd[,8]<600),]
##########################################

# models with only manipulated fctors
# PB, ffd
print(lmer(log(ffd) ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4), cor=FALSE)
print(lmer(log(ffd) ~ c2 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4 & c1=='hi'), cor=FALSE)
print(lmer(log(ffd) ~ c2 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4 & c1=='lo'), cor=FALSE)

# PB, gd
print(lmer(log(gd) ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4), cor=FALSE)
print(lmer(log(gd) ~ c2 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4 & c1=='hi'), cor=FALSE)
print(lmer(log(gd) ~ c2 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4 & c1=='lo'), cor=FALSE)

# PoF, ffd
print(lmer(log(ffd) ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==2), cor=FALSE)

# PoF, gd
print(lmer(log(gd) ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==2), cor=FALSE)
##########################################

#09feb25, first fix landing pos

print(lmer(fst_lp ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==2), cor=FALSE)
print(lmer(fst_lp ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=gd.ffd, subset=region==4), cor=FALSE)


#09feb25, skipping rate

gd.ffd$id = as.numeric(gd.ffd$id)
gd.ffd$sn = as.numeric(gd.ffd$sn)
idx = which(gd.ffd$region==1) # assume the first region is always fixated

skp = gd.ffd[idx,]

skp$fix_n0 = ifelse( gd.ffd$id[idx+1]==gd.ffd$id[idx] & gd.ffd$sn[idx+1]==gd.ffd$sn[idx] & gd.ffd$region[idx+1]==2, 1, 0)
skp$fix_n1 = ifelse((gd.ffd$id[idx+1]==gd.ffd$id[idx] & gd.ffd$sn[idx+1]==gd.ffd$sn[idx] & gd.ffd$region[idx+1]==3) | 
                    (gd.ffd$id[idx+2]==gd.ffd$id[idx] & gd.ffd$sn[idx+2]==gd.ffd$sn[idx] & gd.ffd$region[idx+2]==3), 1, 0)
skp$fix_n2 = ifelse((gd.ffd$id[idx+1]==gd.ffd$id[idx] & gd.ffd$sn[idx+1]==gd.ffd$sn[idx] & gd.ffd$region[idx+1]==4) | 
                    (gd.ffd$id[idx+2]==gd.ffd$id[idx] & gd.ffd$sn[idx+2]==gd.ffd$sn[idx] & gd.ffd$region[idx+2]==4) |
                    (gd.ffd$id[idx+3]==gd.ffd$id[idx] & gd.ffd$sn[idx+3]==gd.ffd$sn[idx] & gd.ffd$region[idx+3]==4), 1, 0)


print(glmer(fix_n1 ~ c2*freq1 + (1|n.0) + (1|n.1) + (1|n.2) + (1|id), data=skp, family=binomial), cor=FALSE)

a = table(skp$cond, skp$fix_n0)
for (n in 1:8)
print(a[n]/(a[n+8]+a[n]))

tapply(skp$fix_n0, list(skp$id, skp$cond), function(x) mean(x, na.rm=TRUE))
tapply(skp$fix_n1, list(skp$id, skp$cond), function(x) mean(x, na.rm=TRUE))
tapply(skp$fix_n2, list(skp$id, skp$cond), function(x) mean(x, na.rm=TRUE))