#Example - Completely Randomised Design 
# Testing hypothesis in one-way ANOVA
library(car)

#Set working directory
setwd("~/Dropbox/ISI SCB/Data sets")

# read in csv file
fb <-read.csv("fabric.csv", TRUE)

Time <- fb$BurnTime
Fabric <-fb$Fabric

#Display the data in boxplots
boxplot(Time~Fabric)

#One-way ANOVA
fab.aov<- aov(Time~Fabric)
summary(fab.aov)


#Multiple Comparisons
fab.tukey <- TukeyHSD(fab.aov)
fab.tukey
plot(fab.tukey)



#residuals and fitted values of model
res  <- fab.aov$residuals
fit<- fab.aov$fitted.values
#residual plot to assess constant variance assumption
plot(fit,res, main="Residual Plot",xlab= "Fitted Values", ylab="Residuals")

#Test for constant variance 
leveneTest(fab.aov)

#assess normality assumption
qqnorm(res, ylab = "Residuals Sample Quantiles")
shapiro.test(res)

hist(res)

#Remove outlier
fbb =rbind(fb[1:7,],fb[9:20,])
Time2 <- fbb$BurnTime
Fabric2 <-fbb$Fabric
boxplot(Time2~Fabric2)

#Analysis without outlier
#One-way ANOVA
fabb.aov<- aov(Time2~Fabric2)
summary(fabb.aov)

#Multiple Comparisons
fabb.tukey <- TukeyHSD(fabb.aov)
fabb.tukey
plot(fabb.tukey)

#residuals and fitted values of model
res  <- fabb.aov$residuals
fit <- fabb.aov$fitted.values

#residual plot to assess constant variance assumption
plot(fit,res, main="Residual Plot",xlab= "Fitted Values", ylab="Residuals")

#Test for constant variance 
leveneTest(fabb.aov)


#assess normality assumption 
qqnorm(res, ylab = "Residuals Sample Quantiles")
shapiro.test(res)

hist(res)