### Team 36

##########
### STEP 0
##########
load("AllData.Rdata")
##########
### STEP 1
##########

### 1. Difficulty of the task. The levels are
### "Easy scrabble" and "Hard scrabble".
### 2. Number of correctly generated words that are unique.
### 3. H0: The number of correctly generated words doesn't depend on the difficulty of the task.
### H1: The number of correctly generated words depends on the difficulty of the task.

##########
### STEP 2
##########

### creating the subset we will use
participants2018 <- allData[allData$SubjectNr >= 301 & allData$SubjectNr <= 348,]
dualTaskSet <- participants2018[participants2018$partOfExperiment == "dualTask",]
dualTaskSet <- droplevels(dualTaskSet)
levels(dualTaskSet$partOfExperiment)

### end of trials
library(plyr)
allDataEndOfTrialsDualTaskOnly <-
  ddply(dualTaskSet, .(SubjectNr, TrialNumber, scrabbleCondition), summarize,
                       nrCorrectScrabbleWords = max(nrCorrectScrabbleWords)
)

### avg number of scrabble words
withoutNeutral <-
  allDataEndOfTrialsDualTaskOnly[
    allDataEndOfTrialsDualTaskOnly$scrabbleCondition != "neutral",
    ]
averageNrScrabbleWords <-
  with(withoutNeutral,
       aggregate(nrCorrectScrabbleWords,
                 list(SubjectNr=SubjectNr,scrabbleCondition=scrabbleCondition),
                 mean))
colnames(averageNrScrabbleWords)[3] <- "averageScrabbleWords"

##########
### STEP 3
##########

levels(averageNrScrabbleWords$scrabbleCondition)
averageNrScrabbleWords2 <- droplevels(averageNrScrabbleWords)

library(ggplot2)

my_plot <- ggplot(averageNrScrabbleWords2, aes(x=scrabbleCondition, y=averageScrabbleWords)) +
  geom_boxplot(stat = "boxplot")
my_plot

scrabbleConditionMeans <- with(averageNrScrabbleWords2, tapply(averageScrabbleWords, list(scrabbleCondition),mean))

barplot(height=scrabbleConditionMeans, main="Average Number of Scrabble Words Generated", ylim=c(0,12))

###
### So far, the pattern we observe in our plots is in line with our hypothesis.
### We expected the easy condition to be larger than the hard condition and this
### is indeed the case.
###
### A within-subjects design has been used.
###

source("usefulFunctions.R")
summaryScrabble <- summarySEwithin(averageNrScrabbleWords2,
                                   measurevar="averageScrabbleWords", withinvars=c("scrabbleCondition"))
print(summaryScrabble)

library(ggplot2)
g1 <- ggplot(summaryScrabble, aes(x=scrabbleCondition, y=averageScrabbleWords)) +
  geom_bar(stat="identity") +
  geom_errorbar(aes(ymin=averageScrabbleWords-sd, ymax=averageScrabbleWords+sd), width=.2) +
  ylab("Average Number of Words") +
  xlab("Task Difficulty") +
  ggtitle("Standard Deviation") +
  scale_y_continuous(limits = c(0, 20))

g2 <- ggplot(summaryScrabble, aes(x=scrabbleCondition, y=averageScrabbleWords)) +
  geom_bar(stat="identity") +
  geom_errorbar(aes(ymin=averageScrabbleWords-se, ymax=averageScrabbleWords+se), width=.2) +
  ylab("Average Number of Words") +
  xlab("Task Difficulty") +
  ggtitle("Standard Errors") +
  scale_y_continuous(limits = c(0, 20))

g3 <- ggplot(summaryScrabble, aes(x=scrabbleCondition, y=averageScrabbleWords)) +
  geom_bar(stat="identity") +
  geom_errorbar(aes(ymin=averageScrabbleWords-ci, ymax=averageScrabbleWords+ci), width=.2) +
  ylab("Average Number of Words") +
  xlab("Task Difficulty") +
  ggtitle("Confidence interval") +
  scale_y_continuous(limits = c(0, 20))


library(cowplot)
print(plot_grid(g1,g2,g3, ncol = 3))

###
### Since the 95% CI bars do not overlap, we expect that in the experiment
### the two conditions will differ in line with our hypothesis H1.
###

##########
### STEP 4
##########

summaryScrabble$lower <- summaryScrabble$averageScrabbleWords - summaryScrabble$ci
summaryScrabble$upper <- summaryScrabble$averageScrabbleWords + summaryScrabble$ci

##########
### STEP 5
##########

with(averageNrScrabbleWords2,t.test(averageScrabbleWords~ scrabbleCondition,paired=TRUE))