---
title: "ESM6_Robert_et_al2023_TrainingChoicesAnalysis"
author: "Théo Robert, Vivek Nityananda"
date: "2023-09-22"
output: html_document
---

```{r setup, include= FALSE}

rm(list=ls())
library(tidyverse)
library(glmmTMB)
library(DHARMa)

#Loading data
trainChoice=read.csv("C:\\Users\\nvn6\\OneDrive - Newcastle University\\Documents\\LearningAttentionPaper\\Revision\\beeTrainingChoices.csv")

choiceSum =trainChoice %>%
  group_by(trainingStage,experiment,flowerColour) %>% 
  summarise_at(vars(rewardChoice), list(name = sum))

```

```{r modelSelection, include= TRUE}
#model selection
m=glmmTMB(rewardChoice~(1|bee),data=trainChoice,family=binomial (link = "logit"))
m1=glmmTMB(rewardChoice~flowerColour+(1|bee),data=trainChoice,family=binomial (link = "logit"))
anova(m1,m)
#model m1 is significantly different

m2=glmmTMB(rewardChoice~experiment+flowerColour+(1|bee),data=trainChoice,family=binomial (link = "logit"))
anova(m2,m1)
#model m2 is not significantly different, m2 is the correct model. This is also supported by the AIC

m3=glmmTMB(rewardChoice~trainingStage+flowerColour+(1|bee),data=trainChoice,family=binomial (link = "logit"))
anova(m3,m1)

#Details of the selected model
SimOutputMTot_NullFirst=simulateResiduals(fittedModel=m3, plot = T)
testDispersion(SimOutputMTot_NullFirst)
summary(m3)


```