---
title: "Statistiques descriptives de l'expérience"
author: "Adrien Coiffard"
date: "09/05/2022"
output: html_document
---

```{r setup, include=FALSE}
rm(list = ls()) 
library(tidyverse)
library(cowplot)

#knitr::opts_knit$set(root.dir = '/tmp')
knitr::opts_chunk$set(echo = FALSE) ## par défaut echo = FALSE
```

## Import des données

On souhaite comparer les caractéristiques socio demographiques de deux échantillons de sujets. 

Pour cela on importe tout d'abord les données au format .csv .

```{r, echo=TRUE}
df <- read.csv("data_StatsDesc.csv")
```


## Statistiques de base (ensemble de l'échantillon)

```{r}
attach(df)

#create new summary function
mySummary <- function(vector){
  results <- c(summary(vector), 'Std. Dev' = sd(vector, na.rm=T))
  results <- round(results, 2)
  return(results)
}

my_ppt_theme <- theme_bw() +
  theme(axis.text.x = element_text(size=12),
        axis.text.y = element_text(size=12),
        axis.title.x = element_text(size=14),
        axis.title.y = element_text(size=14),
        legend.text = element_text(size=12),
        legend.title = element_text(size=14),
        text = element_text(size=16), 
        plot.title = element_text(hjust = 0.5),
        plot.margin = unit(c(0.5,0.1,0.2,0.2), "cm"))
theme_set(my_ppt_theme)
```

Réalisons tout d'abord une série de statistiques descriptives sur l'ensemble de l'échantillon. 


*Question : Was it easy to accomplish the exercice ? (From 0: not at all, to 10: very easy)*

```{r easybid}
mySummary(easyBid)

```
*Questions : Are you a person who is generally risk-taking or do you try to avoid taking risks as much as*
*possible? (From 0: avoid taking risks as much as possible, to 10: very comfortable with the idea*
*of taking risks)*

```{r risk}
mySummary(risk)
```


## Graphiques (par traitements)

```{r}
detach(df)
```

Maintenant regardons graphiquement s'il y a des différences entre les deux traitements (lignes discontinues verticales = moyennes par traitements). 

<br />

*Question : Was it easy to accomplish the exercice ? From 0: not at all, to 10: very easy*

```{r}
# Treatment 1
p1 <- df %>% filter(treatment==0) %>%
ggplot(aes(x=as.factor(easyBid)))+
  geom_bar(aes(y = ..count..),fill="darkblue", width = 0.5) +
  geom_vline(aes(xintercept=mean(easyBid)),size=0.9, color="darkblue",linetype = "dashed")+
  scale_y_continuous(breaks=seq(0,40,10),limits = c(0,40))+
  labs(title="Treatment 1", 
         x="", y = "Nb of Subjects")

# Treatment 2
p2 <- df %>% filter(treatment==1) %>%
ggplot(aes(x=as.factor(easyBid)))+
  geom_bar(aes(y = ..count..),fill="orange2", width = 0.5) +
  geom_vline(aes(xintercept=mean(easyBid)),size=0.9, color="orange2",linetype = "dashed")+
  scale_y_continuous(breaks=seq(0,40,10),limits = c(0,40))+
  labs(title="Treatment 2", 
         x="", y = "")

plot_grid(p1, p2, ncol=2)

# ggplot(df, aes(x=as.factor(easyBid), fill=as.factor(treatment)))+
#   geom_bar(aes(y = (..count..)/sum(..count..)), position="dodge") + 
#   ylab('Percent of participants, %') +
#   scale_y_continuous(labels=scales::percent)

```

Ils semblerait que les sujets aient éprouvé un niveau de difficulté semblable dans les deux traitements. On peut tester cette hypothèse en réalisant un Kruskal–Wallis test :

```{r}
kruskal.test(easyBid ~ treatment, data = df)
```

<br />

*Question : Are you a person who is generally risk-taking or do you try to avoid taking risks as much as*
*possible? (From 0 "avoid taking risks as much as possible", to 10 "very comfortable with the idea*
*of taking risks")*

```{r}
# Treatment 1
p1 <- df %>% filter(treatment==0) %>%
ggplot(aes(x=as.factor(risk)))+
  geom_bar(aes(y = ..count..),fill="darkblue", width = 0.5) +
  geom_vline(aes(xintercept=mean(risk)),size=0.9, color="darkblue",linetype = "dashed")+
  scale_y_continuous(breaks=seq(0,40,10),limits = c(0,40))+
  labs(title="Treatment 1", 
         x="", y = "Nb of Subjects")

# Treatment 2
p2 <- df %>% filter(treatment==1) %>%
ggplot(aes(x=as.factor(risk)))+
  geom_bar(aes(y = ..count..),fill="orange2", width = 0.5) +
  geom_vline(aes(xintercept=mean(risk)),size=0.9, color="orange2",linetype = "dashed")+
  scale_y_continuous(breaks=seq(0,40,10),limits = c(0,40))+
  labs(title="Treatment 2", 
         x="", y = "")

plot_grid(p1, p2, ncol=2)
```

Ils semblerait qu'il n'y ait pas non plus de différence significative entre les deux traitements sur l'aversion au risque. On peut à nouveau tester cette hypothèse en réalisant un Kruskal–Wallis test :

```{r}
kruskal.test(risk ~ treatment, data = df)
```