##########################
## CQS Summer Institute ##
##    Biostatistics 1   ##
##    Tatsuki Koyama    ##
##########################

## MouseWeight.xls has the data from an experiment. 
## Mice 11~20 are in 'Uninfected' group, and mice 21~30 are in 'Infected'
## group. 

## 1. Data cleaning ##
## This data sheet requires some cleaning. 

## First, we need this file as a .csv (text) file. Save it as
## 'MouseWeight.csv'. (Some of the information gets lost.)

## Second, Mean and SD are not a part of the data. They do not belong here with
## the data. Delete them. 

## Third, each mouse needs 'Group' variable. Fill in the column. (There is an
## easy way to do this in R, but copy-pasting is ok for now.)

## Read in the data.
mouseWeight <- read.csv('MouseWeight.csv', sep=',')

## Data summary.
## Let's get some summaries of the data.
## groupSum() function is useful. 
source('http://biostat.mc.vanderbilt.edu/wiki/pub/Main/TatsukiRcode/RFunctions071817.R')
groupSum(mouseWeight$Baseline, mouseWeight$Group)

## Let's look at the data. 
source('http://biostat.mc.vanderbilt.edu/wiki/pub/Main/TatsukiRcode/RFunctions0.R')
## Some documentation on http://biostat.mc.vanderbilt.edu/wiki/Main/TatsukiRcode
tplot(Baseline ~ Group, data=mouseWeight)
## Try different options...
tplot(Baseline ~ Group, data=mouseWeight, type='db', show.n=TRUE, ylim=c(10,18))
## What do you notice?

## Q1 Test whether the baseline weight is different (on average) between the groups.
## Write the hypotheses.
t.test(mouseWeight$Baseline ~ mouseWeight$Group)
t.test(Baseline ~ Group, data=mouseWeight)

wilcox.test(mouseWeight$Baseline ~ mouseWeight$Group)
## What can you say?

## Q1 Test whether 'infected' mice, on average, lose more weight by Day 1. 
## Q1a. Write the hypotheses.
##
## Q1b. Create a column 'change1', which is weight change from Baseline to Day 1.
mouseWeight$change1 <- mouseWeight$Day1 - mouseWeight$Baseline
    groupSum(mouseWeight$change1, mouseWeight$Group, Combined=TRUE, Test=TRUE)
    tplot(change1 ~ Group, data=mouseWeight, show.n=TRUE, median.line=TRUE)
        title('Weight change', adj=0)
t.test(change1 ~ Group, data=mouseWeight, alt='less')

## to get 90% confidence interval...
t.test(change1 ~ Group, data=mouseWeight, alt='two', conf=0.90)

## Q1c. How about using the fold-change?
mouseWeight$foldChange1 <- mouseWeight$Day1 / mouseWeight$Baseline
    groupSum(mouseWeight$foldChange1, mouseWeight$Group, Combined=TRUE, Test=TRUE)
    tplot(foldChange1 ~ Group, data=mouseWeight, show.n=TRUE, median.line=TRUE)
        title('Weight fold-change', adj=0)
t.test(foldChange1 ~ Group, data=mouseWeight, alt='less')

## Q2
## The following data were reported in 
## Preliminary Report: Findings from teh Aspirin Component of the Ongoing
## Physicians' Health Study. N. Engl. J. Med. 318:262-264, 1988. 
tab <- data.frame(rbind( c(18+171, 10845), c(5+99, 10933) ))
    row.names(tab) <- c('Placebo','Aspirin')
    names(tab) <- c('Fatal and Non-fatal Attack', 'No Attack')

## Q2a Find the proportions of heart attacks for each treatment group.
    ## tab[,1] / (tab[,1] + tab[,2])
    ## ## 189 / 11037 and 104 / 11034

## Q2b Find the odds of heart attacks for each treatment group.
    ## tab[,1] / tab[,2]

## Q2c Find the odds ratio and interpret the number. 
    ## odds <- tab[,1] / tab[,2]
    ## odds.ratio <- odds[1] / odds[2]
    ## ## The odds of heart attack is 1.83 times as high for those taking
    ## placebo than those taking aspirin.

## Conduct a Fisher's exact test to test the odds are the same.
    ## fisher.test(tab)
## Conduct a chi-sq test to test P[Heart Attack] are the same.
    ## chisq.test(tab, correc=FALSE)