---
title: "Connect Project"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
# Load RColorBrewer
# install.packages("RColorBrewer")
library(RColorBrewer)

# Define colour palettes for plots below
coul3 <- brewer.pal(3, "RdYlBu") # Using RdYlBu range to generate 3 colour palette: https://colorbrewer2.org/#type=diverging&scheme=RdYlBu&n=5

```

Note that the `echo = FALSE` parameter was added to the code chunk to prevent printing of the R code that generated the plot.

### To Do List

## Upload Data
```{r Data Upload}
connect_data = read.csv("./data/connectDATA.csv")
```

## Summary of Data
Data summary/visualisation with subsetting:
  - RH: display simple summary of data (bar/pie chart) to Q25/26, Q3
  
```{r Frequencies}  
#Frequencies#
  
Q25_frequencies <- table(connect_data$Q25)
Q25_frequencies
Q26_freq <- table(connect_data$Q26)
Q26_freq
Q3_freq <- table(connect_data$Q3)
Q3_freq

#test3 = as.factor(connect_data$Q3, levels = c(1, 2, 3, 4, 5), labels = c("Worldviews", "Religion", "Theology", "Ethics", "Philosophy"))
```

```{r Q25 bar/pie}
pie(Q25_frequencies, labels = c("Maybe", "No", "Yes"))
pie(Q25_frequencies, labels = c("Maybe", "No", "Yes"), col = coul3)

```

```{r Q26 bar/pie}

Q26_data <- read.csv("./data/Q26_data.csv")

Q26_freq_data <- data.frame(c("Other Priorities", "Lack Subject Knowledge", "Lack Confidence", "Current Syllabus", "Pupil Disinterest", "Department Head", "Available Work Schemes", "Unavailable Resources", "Uncertain of Pedagogical Approach"), c(table(Q26_data[,2]) [names(table(Q26_data[,2])) == "TRUE"],
                table(Q26_data[,3]) [names(table(Q26_data[,3])) == "TRUE"],
                table(Q26_data[,4]) [names(table(Q26_data[,4])) == "TRUE"],
                table(Q26_data[,5]) [names(table(Q26_data[,5])) == "TRUE"],
                table(Q26_data[,6]) [names(table(Q26_data[,6])) == "TRUE"],
                table(Q26_data[,7]) [names(table(Q26_data[,7])) == "TRUE"],
                table(Q26_data[,8]) [names(table(Q26_data[,8])) == "TRUE"],
                table(Q26_data[,9]) [names(table(Q26_data[,9])) == "TRUE"],
                table(Q26_data[,10]) [names(table(Q26_data[,10])) == "TRUE"]))

head(Q26_freq_data)
names(Q26_freq_data)[1] <- "Reasons"
names(Q26_freq_data)[2] <- "Frequency"
head(Q26_freq_data)
pie(Q26_freq_data$Frequency, labels = c("Other Priorities", "Lack Subject Knowledge", "Lack Confidence", "Current Syllabus", "Pupil Disinterest", "Department Head", "Available Work Schemes", "Unavailable Resources", "Uncertain of Pedagogical Approach"))

# Bar graph tidier


```

pie(Q26_freq) 
  #very messy as a pie chart - split by type? Or is it important to see crossover
  
  Could potentially see crossover with crosstabs by type (since response is now binary variable T/F), maybe chi square; perhaps just descriptives
  

```{r Q3 bar/pie}

Q3_data <- read.csv("./data/Q3.csv")

#head(Q3_data)
#table(Q3_data [,3:7])
#pie(table(Q3_data [,3:7]))

Q3_data2 <- Q3_data[ ,3:7]
#head(Q3_data2)
#table(Q3_data2)
#table(Q3_data2[,1])

### want to take only the count of "True" (1) in each column. Then pie chart of the frequencies

#Q3_data3 <- read.csv("~/Documents/Github/re_connect_survey/data/Q3 copydata.csv")

#table(Q3_data3)
  #count(Q3_data3, 1)
#table(Q3_data3) [names(table(Q3_data3)) == 1]
#table(Q3_data3)

table(Q3_data2[,1]) [names(table(Q3_data2[,1])) == "TRUE"]

test2 <- data.frame(c("Worldviews", "Religion", "Theology", "Ethics", "Philosophy"), c(table(Q3_data2[,1]) [names(table(Q3_data2[,1])) == "TRUE"],
                table(Q3_data2[,2]) [names(table(Q3_data2[,2])) == "TRUE"],
                table(Q3_data2[,3]) [names(table(Q3_data2[,3])) == "TRUE"],
                table(Q3_data2[,4]) [names(table(Q3_data2[,4])) == "TRUE"],
                table(Q3_data2[,5]) [names(table(Q3_data2[,5])) == "TRUE"]))
head(test2)
names(test2)[1] <- "Subject"
names(test2)[2] <- "Frequency"
head(test2)
pie(test2$Frequency, labels = c("Worldviews", "Religion", "Theology", "Ethics", "Philosophy"))

# JK note on Q3: consider here whether to use alternative forms of visualiation to reflect the overlaps when respondents picked multiple categories in responses

```

xtabs(Frequency ~ Subject, test2)

pie(Q3_freq)
  #also not optimal as pie...perhaps bar

  #sum(Q3_data2)
  
Q3_1factor = as.factor(Q3_data2$Religion)
table(Q3_1factor)
   #count(Q3_1factor, "TRUE")

  #test = replace(Q3_1factor, "TRUE", 1)
  #test
  #Q3_1factor

  - RH: display summaries of responses to key questions for Q22 (syllabus evaluation), Q23, Q24, Q25, Q26, Q27, with subsetting by:
    - Q8 (school type)
    - Q9 (school size)
    - Q10 (school location)
    - Q1 (grade level) + Q35 (teaching role) + +Q5 (teaching proportion) Q2 (tenure) + and Q3 (subjects taught), + Q6/Q7 (management)
    - Q12-14 (school's official religion) / Q15-16 (school's informal religion)
    - Q21 (respondent personal religious background)
    - Q4 (teacher's degree subject)
    - Q18 (respondent gender)
    - Q19 (respondent ethnic self-desc)
    
```{r Plots}
library(ggplot2)

# Q22
testplot <- 
# Q23

# Q24

# Q25

# Q26

# Q27
```

## Correlation testing:
  - RH: test for correlation between "social issue" box ticked on Q20 and responses to Q22, Q23, Q27 
  
  - Make Q20 a factor with 14 levels
  - Collapse 2 Q22 columns into one mean for analyses
  - Analyse 1 way anova Q20 (14 levels) by Q22; Q23[1-2]; Q27[1-7] 
  - 1 way within subjects?? Though not all participants ticked every box... Would it then be best to separate them out and do 14 separate analyses with bonferroni correction due to the multiple tests? - could then be 14 different t tests based on whether they ticked each one as important or not... Many analyses but that may be the most straightforward way to go. Factorial mixed ANOVA? 14 predictors, each with 2 levels (yes/no)??
  - 14 predictors, within subjects, 2 levels (yes/no). DV as responses to questions. Q22 would be a factorial between subjects (only 1 option on IVs) ANOVA. Qs 23, 27 would be factorial between subjects MANOVA 
  

```{r Analyses 1 - As Factor}
social_issues_data <- read.csv("./data/Q20_data.csv")
head(social_issues_data)

# All 14 as factors, with 2 levels: 1=YES, 2=NO

social_issues_data$brexit <- factor(social_issues_data$brexit, levels = c(1, 2), labels = c("Yes", "No"))
class(social_issues_data$brexit)

#social_issues_data[ ,4:5] <- factor(social_issues_data[ ,4:5], levels = c(1, 2), labels = c("Yes", "No"))
  #Did not work; made 2 columns "NA" so am going through to make factors     individually

  ### OR ###

#social_issues_data[ ,4:5] <- lapply(social_issues_data[ ,4:5], factor(social_issues_data[ ,4:5], levels = c(1, 2), labels = c("Yes", "No")))

social_issues_data$economy <- factor(social_issues_data$economy, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$immigration <- factor(social_issues_data$immigration, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$crime <- factor(social_issues_data$crime, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$health <- factor(social_issues_data$health, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$education <- factor(social_issues_data$education, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$housing <- factor(social_issues_data$housing, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$welfare <- factor(social_issues_data$welfare, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$defence <- factor(social_issues_data$defence, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$environment <- factor(social_issues_data$environment, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$tax <- factor(social_issues_data$tax, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$pensions <- factor(social_issues_data$pensions, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$family.life <- factor(social_issues_data$family.life, levels = c(1, 2), labels = c("Yes", "No"))

social_issues_data$transport <- factor(social_issues_data$transport, levels = c(1, 2), labels = c("Yes", "No"))

```

``` {r Analyses 2 - ANOVA and MANOVA}
## Q22; Q23[1-2]; Q27[1-7]
  #Q22_average
  #Q23_1, Q23_2
  #Q27_1 - Q27_7
#t.test to see if difference in one variable - Q22_average
hist(social_issues_data$Q22_average)
t.test(Q22_average~brexit, data = social_issues_data, paired = FALSE)
  #no significant difference between scores on Q22, and whether they thought      brexit was important


Q_22test <- aov(Q22_average ~ brexit + economy + immigration + crime + health + education + housing + welfare + defence + environment + tax + pensions + family.life + transport, data = social_issues_data)
summary(Q_22test)
  #no significant different between scores on Q22 and their opinion on social     issues

Q_23test <- manova(cbind(Q23_1, Q23_2) ~ brexit + economy + immigration + crime + health + education + housing + welfare + defence + environment + tax + pensions + family.life + transport, data = social_issues_data)
summary(Q_23test)
  #significant difference between scores on Q23 with economy, health, and         environment

econ <- aggregate(cbind(Q23_1, Q23_2) ~ economy, data = social_issues_data, FUN = mean)
health <- aggregate(cbind(Q23_1, Q23_2) ~ health, data = social_issues_data, FUN = mean)
env <- aggregate(cbind(Q23_1, Q23_2) ~ environment, data = social_issues_data, FUN = mean)


#SORT OUT MEANS FOR THIS -- interesting pattern viewed with means

Q_27test <- manova(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ brexit + economy + immigration + crime + health + education + housing + welfare + defence + environment + tax + pensions + family.life + transport, data = social_issues_data)
summary(Q_27test)
  #No significant difference in responses to Q27 based on what they considered   important
```
  - RH: test for correlation between responses to religion questions: Q12-14, Q15-16 and Q21 and responses to Q22, Q23, Q27,  [Q24, Q25, Q30]
  
``` {r Analyses based on religious affiliation}
religion_affiliation_data <- read.csv("./data/Religious affiliation data.csv")
head(religion_affiliation_data)

## Q12-14, with Q22, Q23, Q27
  # Q12 is binary, 1st test whether difference in answers based on whether the school has formal religious character or not (similar ANOVA/MANOVA as the questions above)
religion_affiliation_data$Q12 <- factor(religion_affiliation_data$Q12, levels = c("No", "Yes"), labels = c("No", "Yes"))

  ## Q22
formal_affiliation_test_Q22 <- t.test(Q22_average ~ Q12, data = religion_affiliation_data, paired = FALSE)
formal_affiliation_test_Q22

  ## Q23
formal_affiliation_test_Q23 <- manova(cbind(Q23_1, Q23_2) ~ Q12, data = religion_affiliation_data)
summary(formal_affiliation_test_Q23)

  ## Q27
formal_affiliation_test_Q27 <- manova(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ Q12, data = religion_affiliation_data)
summary(formal_affiliation_test_Q27)
  
# Then, if there is (or can anyway), explore only the "Yes" data, and see if there is a difference in answers based on the specific religious character -- Q13
  # first subset the data
Q13_data <- religion_affiliation_data[religion_affiliation_data$Q12 == "Yes", ]
head(Q13_data)

  # then analyze based on specific one
Q13_data$Q13_recode <- factor(Q13_data$Q13_recode, levels = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10), labels = c("Church of England", "Roman Catholic", "Methodist", "Other Christian", "Jewish", "Muslim", "Sikh", "Hindu", "Multi-Faith", "None of the above"))

# Test with only included levels
Q13_data$Q13_recode <- factor(Q13_data$Q13_recode, levels = c(1, 2, 4, 6), labels = c("Church of England", "Roman Catholic", "Other Christian", "Muslim"))
# No change with this one...still nonsignificant difference

  # Q22
hist(Q13_data$Q22_average)
specific_affiliation_test <- aov(Q22_average ~ Q13_recode, data = Q13_data)
typeof(Q13_data$Q13_recode)
summary(specific_affiliation_test)

  # Q23
specific_affiliation_test_Q23 <- manova(cbind(Q23_1, Q23_2) ~ Q13_recode, data = Q13_data)
summary(specific_affiliation_test_Q23)

  # Q27
specific_affiliation_test_Q27 <- manova(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ Q13_recode, data = Q13_data)
summary(specific_affiliation_test_Q27)

## Q15-16 with Q22, Q23, Q27
  # Q15 is binary; 1st test whether difference in answers based on whether the school has an informal religious character or not. Q16 provides further detail and can be explored

religion_affiliation_data$Q15 <- factor(religion_affiliation_data$Q15, levels = c("No", "Yes"), labels = c("No", "Yes"))

  ## Q22
informal_affiliation_test_Q22 <- t.test(Q22_average ~ Q15, data = religion_affiliation_data, paired = FALSE)
informal_affiliation_test_Q22

  ## Q23
informal_affiliation_test_Q23 <- manova(cbind(Q23_1, Q23_2) ~ Q15, data = religion_affiliation_data)
summary(informal_affiliation_test_Q23)

  ## Q27
informal_affiliation_test_Q27 <- manova(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ Q15, data = religion_affiliation_data)
summary(informal_affiliation_test_Q27)

```

```{r Analyses based on personal religious affiliation}
## Q21 with Q22, Q23, Q27
  # Q21 is personal religious affiliation. This may be more tricky as it is a free answer...but can code the type of religious affiliation and test that way? -- would be chi-square or some sort of non-para analysis due to the small number of respondents who answered this

personal_religious_affiliation_data <- read.csv("./data/Personal religious affiliation data.csv")

head(personal_religious_affiliation_data)

personal_religious_affiliation_data$Q21_binaryrecode <- factor(personal_religious_affiliation_data$Q21_binaryrecode, levels = c(1, 2), labels = c("none", "answered"))

  ## Q22
personal_religious_affiliation_test_Q22 <- t.test(Q22_avg ~ Q21_binaryrecode, data = personal_religious_affiliation_data, paired = FALSE)
personal_religious_affiliation_test_Q22

  ## Q23
personal_religious_affiliation_test_Q23 <- manova(cbind(Q23_1, Q23_2) ~ Q21_binaryrecode, data = personal_religious_affiliation_data)
summary(personal_religious_affiliation_test_Q23)

  ## Q27
personal_religious_affiliation_test_Q27 <- manova(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ Q21_binaryrecode, data = personal_religious_affiliation_data)
summary(personal_religious_affiliation_test_Q27)

  # Significant difference between answers to Q27 and whether participants indicated a personal religious affiliation -- with the small sample size it may be easier to visualize the differences here based on freeform answer

head(personal_religious_affiliation_data)
personal_religious_affiliation_means <- aggregate(cbind(Q27_1, Q27_2, Q27_3, Q27_4, Q27_5, Q27_6, Q27_7) ~ Q21_binaryrecode, data = personal_religious_affiliation_data, FUN = mean)
personal_religious_affiliation_means

## In viewing the means, it is likely the significant difference viewed in the above MANOVA is within Q27_7, with those who indicated having a personal religious affiliation reporting lower scores (M = 2.94) than those who did not answer or indicated they had no religious affiliation (M = 3.83). This makes sense as a higher score indicates they disagree that they know about "how they put their beliefs about the climate/biodiversity crisis into practice"
  # Also a slight difference in Q27_3 with those indicating having a personal      religious affiliation reporting slightly lower scores (M = 3.24) than those who   did not answer to indicated they had no religious affiliation (M = 3.76)
```