library(lme4)
library(mgcv)
library(lattice)

d <- read.table("tutorial.txt", head=TRUE)

#### #### #### ####

m1 <- lmer(duration~mary_context +(1|speakerid), data=d)
m2 <- lmer(duration~mary_context+amiFreq +(1|speakerid), data=d)
summary(m1)
summary(m2)

anova(m1,m2)

m1b <- lmer(duration~mary_context + (1+mary_context | speakerid), data=d)
summary(m1b)
anova(m1, m1b)
# show the effect of individuals
print(dotplot(ranef(m1b,cond=TRUE),  scales = list(x = list(relation = 'free')))[["speakerid"]])

# consider only subset of french native speakers:
m1bfrench <- lmer(duration~mary_context + (1+mary_context | speakerid), data=subset(d, native_language=="french"))
print(dotplot(ranef(m1bfrench,cond=TRUE),  scales = list(x = list(relation ='free')))[["speakerid"]])

# add out of domain trigram as a predictor
m3<-lmer(duration~mary_context+amiFreq+srpTrigram.OODomain+(1|speakerid),data=d)
summary(m3)
anova(m3,m2)

# look at in domain trigram model
m3in<-lmer(duration~mary_context+amiFreq+srpTrigram.inDomain  +( 1|speakerid), data=d)
summary(m3in)
summary(m3)

par(mfcol=c(2,1))
plot(gam(duration~mary_context + amiFreq + s(srpTrigram.inDomain), data=d))
plot(gam(duration~mary_context + amiFreq + s(srpTrigram.OODomain), data=d))

m3full<-lmer(duration~mary_context+amiFreq+srpTrigram.OODomain+(mary_context+amiFreq+srpTrigram.OODomain+1|speakerid),data=d)
summary(m3full)

m3infull<-lmer(duration~mary_context+amiFreq+srpTrigram.inDomain  +(mary_context+amiFreq+srpTrigram.inDomain  +1|speakerid), data=d)
summary(m3infull)