# Lab Exercise 4 Solutions

#cardat.csv
dat<-read.csv(file.choose())
head(dat)
names(dat)

# 1
library(lawstat)

#Homogeneity of Variance
tapply(dat$purc,dat$type,var)
lawstat::levene.test(dat$purc,dat$type)

#Normality
hist(dat$purc[dat$type=="car"])
hist(dat$purc[dat$type=="truck"])
hist(dat$purc[dat$type=="van"])
boxplot(dat$purc~dat$type)
qqnorm(dat$purc[dat$type=="car"])
qqnorm(dat$purc[dat$type=="truck"])
qqnorm(dat$purc[dat$type=="van"])

tapply(dat$purc,dat$type,shapiro.test)

#Welch F Test
oneway.test(purc~type, data=dat)

# Assumptions are reasonably met, 
# but just in case

#Welch on Trimmed Means
library(WRS2)
t1way(purc~type, tr=.2, data=dat)

# 2
pairwise.t.test(dat$twoyr, dat$type, p.adj="holm")

# 3
long_dat<-reshape(data=dat, direction="long", timevar="time",varying=1:3, v.names='shape',sep="")

library(ez)
long_dat$time<-factor(long_dat$time)
ezANOVA(wid=id,within=time,dv=shape, data=long_dat)

library(nlme)
mod1<-lme(shape~time, random=~1|id, data=long_dat)
anova(mod1)

## shows how to get p values
pvals<-c()
pvals[1]<-t.test(dat$purc,dat$oneyr, paired=TRUE)$p.value
pvals[2]<-t.test(dat$purc,dat$twoyr, paired=TRUE)$p.value
pvals[3]<-t.test(dat$oneyr,dat$twoyr, paired=TRUE)$p.value
pvals
p.adjust(p=pvals,method='holm')

# 4
interaction.plot(x.factor=dat$sex,
    trace.factor=dat$type,response=dat$twoyr)
mod1<-lm(twoyr~sex*type, data=dat)
anova(mod1)

# 5
library(ez)
ezANOVA(data=long_dat,wid=id,dv=shape,within=time,between=sex)
interaction.plot(long_dat$time,long_dat$sex,long_dat$shape)

#Interaction Contrasts
ezANOVA(data=subset(long_dat,time==1|time==2),wid=id,dv=shape,within=time,between=sex)
ezANOVA(data=subset(long_dat,time==1|time==3),wid=id,dv=shape,within=time,between=sex)
ezANOVA(data=subset(long_dat,time==2|time==3),wid=id,dv=shape,within=time,between=sex)

library(nlme)
mod2<-lme(shape~time*sex, random=~1|id, data=long_dat)
anova(mod2)