# Hierarchical models. Estimation only, no NHST testing.

library(lme4)
library(lattice)

# ---- Look at the data ---
data(sleepstudy)
head(sleepstudy)

xyplot(Reaction ~ Days | Subject, data=sleepstudy, type=c("p","r") )

?lmer
?glmer # for logistic, Poisson and other models

# ------ Model fitting ---------
# -- Wrong model which ignores repeated measures
fit0 <- lm(Reaction ~ Days, data=sleepstudy)
summary(fit0)

plot(Reaction ~ Days, data=sleepstudy)
abline(coef(fit0))

# -- Random intercept, fixed slopes (Days)
fit1 <- lmer(Reaction ~ Days + (1|Subject), data=sleepstudy )
summary(fit1) # "general effect" for (Intercept) = 251.4051
ranef(fit1) # matrix of effects added to "general level" (Intercept)

# --- Random intercept, random slopes
fit2 <- lmer(Reaction ~ Days + (1+Days|Subject), data=sleepstudy )
summary(fit2)
ranef(fit2)