Descriptive analysis_How do people feel about AI.Rmd

---
title: "Descriptive analysis"
output:
  word_document: default
  html_document: default
date: "2023-05-10"
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


```{r}
# setwd("..") # Use this line to set the working directory on your machine where the data is located. 
```

```{r}
install.packages("tidyverse")
install.packages("haven")
install.packages("scales")
install.packages("psych")
install.packages("survey")
install.packages("jtools")
install.packages("nFactors")
install.packages("sjPlot") 
install.packages("purrr")
install.packages("broom")
install.packages("labelled")
install.packages("stringr")
install.packages("descr")
```


```{r}
library(tidyverse)
library(haven)
library(scales)
library(psych)
library(survey)
library(jtools)
library(nFactors)
library(sjPlot) 
library(purrr)
library(broom)
library(labelled)
library(stringr)
library(descr)
```

```{r}
# Read dataset
df1 <- read_dta("262400371_Public Attitudes to AI_client version.dta")
```

```{r}
# Prep data
# Remove ppt identifying as 'other' and remove 16 year old 
df1 <- df1 %>%
  filter(is.na(RS_Sex) | RS_Sex != 3) %>%
  filter(CS_Age != "16")

# combine ethnicity groupings
df1 <- df1 %>%
  mutate(RS_Ethnicity_Combined = 
           case_when(
             RS_Ethnicity_Reduced == 1 ~ 1, 
             RS_Ethnicity_Reduced == 2 ~ 2,
             RS_Ethnicity_Reduced == 3 ~ 3,
             RS_Ethnicity_Reduced == 4 | RS_Ethnicity_Reduced == 5 | RS_Ethnicity_Reduced == 6 ~ 4 
           ))

            #"1" = "White",
            # "2" = "Asian or Asian British",
            # "3" = "Black British, Caribbean or African",
            # "4" = "Any other ethnic group"

df1 <- df1 %>%
  mutate(Region_combined = 
           case_when(
             CS_Region_N < 10 ~ 1, #eng
             CS_Region_N == 10 ~ 2, #ni
             CS_Region_N == 11 ~ 3, #scotland
             CS_Region_N == 12 ~ 4 # wales
           ))
```


```{r}
## functions to simplify data tidying

## recoding technology labels
recode_technologies <- function(Technology_specific) {
  recode(Technology_specific,
                  "WEL" = "Welfare eligibility",
                "JOB" =  "Job eligibility",
                "CAN" =  "Cancer risk",
                "LOAN" = "Loan repayment risk", 
                "MP" =  "Mobile phone unlock",
                 "BC" = "Border control",
                 "PS" = "Police surveillance",

                "AW"  = "Autonomous weapons", 
                "DC" =  "Driverless cars",
                "RCA" = "Robotic care assistant",
               "RVC" ="Robotic vacuum cleaner",
                                                                         
               "SMC"=  "Targeted consumer ads",
                "SMP" = "Targeted political ads",
                                                                         
                "VASS"=  "Smart speaker",
                "VAH"=  "Virtual healthcare assistant",
                                 
              "CCR"=   "Climate change simulation",
                "VR"=  "Education simulations")
  
}

## list of demographic variables of interest
demog <- df1 %>%
  remove_val_labels() %>%
  dplyr::select(
  Respondent_ID, PV25_Weight,
  RS_Ethnicity_Combined,
  CS_Age_Band,
  RS_Sex,
  RS_EducationLevel,
  DR_Socio_Eco
  ) %>%
  mutate(RS_Ethnicity_Combined = 
           dplyr::recode(RS_Ethnicity_Combined,
             "1" = "White",
             "2" = "Asian or Asian British",
             "3" = "Black British, Caribbean or African",
             "4" = "Any other ethnic group"
           )) %>%
  mutate(CS_Age_Band = 
           dplyr::recode(CS_Age_Band,
             "1" = "18-24 yrs",
             "2" = "25-34 yrs",
             "3" = "35-44 yrs",
             "4" = "45-54 yrs",
             "5" = "55-64 yrs",
             "6" = "65-74 yrs",
             "7" = "75+ yrs"
           )) %>%
  mutate(RS_Sex = 
           dplyr::recode(RS_Sex,
             "1" = "Male",
             "2" = "Female")) %>%
  mutate(RS_EducationLevel = 
           dplyr::recode(RS_EducationLevel,
             "1" = "Degree level qualification(s)",
             "2" = "Non-degree level qualifications",
             "3" = "No academic or vocational qualifications")) %>%
  mutate(DR_Socio_Eco = 
           dplyr::recode(DR_Socio_Eco,
             "1" = "SEC1, 2",
             "2" = "SEC3",
             "3" = "SEC4",
             "4" = "SEC5",
             "5" = "SEC6, 7",
             "6" = "SEC8",
             "7" = "Students")) %>%
  pivot_longer(names_to = "Q_Demo",
               values_to = "A_Demo",
               RS_Ethnicity_Combined:DR_Socio_Eco)

# list of levels 
demo_list <- c("White",
  "Asian or Asian British",
  "Black British, Caribbean or African", 
  "Any other ethnic group", 
  
  "18-24 yrs",
  "25-34 yrs",
  "35-44 yrs",
  "45-54 yrs",
  "55-64 yrs",
  "65-74 yrs",
  "75+ yrs", 
  
  "Male",
  "Female",
  "Identify in another way",
  
  "Degree level qualification(s)",
  "Non-degree level qualifications",
  "No academic or vocational qualifications",
  
  "SEC1, 2",
  "SEC3",
  "SEC4",
  "SEC5",
  "SEC6, 7",
  "SEC8",
  "Students", 
  
  "Total" )
```




####### Descriptive analysis #########

Awareness of technology
```{r}
# prep data
df2.1 <- df1 %>%
  # remove all labels
  remove_val_labels() %>%
  ## select vars
  select(Respondent_ID, PV25_Weight,
    tidyselect::vars_select(names(df1), starts_with("AWARE_", ignore.case = TRUE))) %>%
  ## pivot longer
  pivot_longer( # convert to long format 
    names_to = "Question_a",
    values_to = "Value_a",
    AWARE_MP:AWARE_VR
  ) 

## create new labelling var for specific technologies
df2.1$Technology_specific <- sub('.*_', '', df2.1$Question_a) # extract everything after the underscore into a new variable by replacing everything before the underscore with blank space.



df2.1 <- df2.1 %>%
  mutate("Value_a" = 
  dplyr::recode(Value_a, 
      "1" = "Yes",
      "2" = "No",
      "3" = "Not sure/prefer not to say",
      "4" = "Not sure/prefer not to say"))

## relabel the codes so they read as words
df2.1 <- df2.1 %>%
  mutate(Technology_specific = 
           recode_technologies(Technology_specific))

## Awareness plot
df2.1 %>%
  drop_na(Value_a) %>%
  group_by(Technology_specific, Value_a) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value_a == "Yes" ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Technology_specific, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Technology_specific, -Sum_f), fill = Value_a)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 12)) +
  labs(title="Awareness of technology",
       subtitle = "'Before today, had you heard of the use of AI technologies for...'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Yes", "Not sure/prefer not to say", "No"), 
                                values = c(
                               "#00A087FF",
                               "#D5DFE8",
                               "#E64B35FF"
                               ),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip()

```

```{r}
## Awareness x demographics of interest 
df2.1.1 <- inner_join(df2.1, demog)

# overall awareness
df2.1.t <- df2.1 %>%
  drop_na(Value_a) %>%
  group_by(Technology_specific, Value_a) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))
  
df2.1.t$A_Demo <- "Total"
df2.1.t$Q_Demo <- "Total"

# awareness by demographics
df2.1.d <- df2.1.1 %>%
  drop_na(Value_a) %>%
  drop_na(A_Demo) %>%
  group_by(Technology_specific, Q_Demo, 
           A_Demo, 
           Value_a) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))

df2.1.1 <- rbind(df2.1.t, df2.1.d)

df2.1.1$A_Demo <- factor(df2.1.1$A_Demo,levels = demo_list)

```

```{r, fig.width=12, fig.height=20}
df2.1.1 %>%
  drop_na(A_Demo) %>%
  ggplot(aes(y=pct, x=A_Demo)) + 
  geom_col(aes(fill = Value_a), width = 0.7) + 
  scale_y_continuous(name = "Percentage (%)", label = percent) +
  theme_light() + 
  facet_wrap(~Technology_specific) + 
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 14)) +
  labs(title="Awareness") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Yes", "Not sure/prefer not to say", "No"), 
                                values = c(
                               "#00A087FF",
                               "#D5DFE8",
                               "#E64B35FF"
                               ),
                    name = "Response") +
  coord_flip()

# awareness x age specific stats
tab <- df2.1.1 %>%
  filter(A_Demo == "75+ yrs") %>%
  filter(Value_a == "Yes")

# mobile phones
# border control
# SMC 

df1 %>%
  drop_na(AWARE_SMC) %>% # change for each use of AI interested in  
  drop_na(CS_Age_Band) %>%
  filter(CS_Age_Band!= 7) %>%
  group_by(AWARE_SMC) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))
```

#######################################################################################################################################

Experience with technology
```{r}
## Experience with technology
df2.2 <- df1 %>%
  # remove all labels
  remove_val_labels() %>%
  ## select vars
  select(Respondent_ID, PV25_Weight, 
    tidyselect::vars_select(names(df1), starts_with("EXP_", ignore.case = TRUE)),) %>%
  ## pivot longer
  pivot_longer( # convert to long format 
    names_to = "Question_e",
    values_to = "Value_e",
    EXP_MP:EXP_VR
  ) 

## create new labelling var for specific technologies
df2.2$Technology_specific <- sub('.*_', '', df2.2$Question_e) # extract everything after the underscore into a new variable by replacing everything before the underscore with blank space.


df2.2 <- df2.2 %>%
  mutate("Value_e" = 
  dplyr::recode(Value_e, 
      "1" = "A lot",
      "2" = "Some",
      "3" = "None",
      "4" = "Not sure/prefer not to say",
      "5" = "Not sure/prefer not to say"))

## relabel the codes so they read as words
df2.2 <- df2.2 %>%
  mutate(Technology_specific = 
           recode_technologies(Technology_specific))

## relevel values
df2.2$Value_e <- factor(df2.2$Value_e,levels = c(   "None",  "Not sure/prefer not to say", "Some", "A lot"))

## Experience plot
df2.2 %>%
  drop_na(Value_e) %>%
  group_by(Technology_specific, Value_e) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value_e == "None"   ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Technology_specific, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Technology_specific, +Sum_f), fill = Value_e)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(
        panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 12)) +
  labs(title="Experience with technology",
       subtitle = "'How much personal experience have you had, if any with...'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("A lot", "Some", "Not sure/prefer not to say", "None" ), 
                                values = c(
                               "#00A087FF",
                               "#91D1C2FF",
                               "#D5DFE8",
                               "#E64B35FF"
                               ),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip()
```

```{r}
## experience x demographics of interest 
df2.2.1 <- inner_join(df2.2, demog)

# overall awareness
df2.2.t <- df2.2 %>%
  drop_na(Value_e) %>%
  group_by(Technology_specific, Value_e) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))
  
df2.2.t$A_Demo <- "Total"
df2.2.t$Q_Demo <- "Total"

# awareness by demographics
df2.2.d <- df2.2.1 %>%
  drop_na(Value_e) %>%
  drop_na(A_Demo) %>%
  group_by(Technology_specific, Q_Demo, 
           A_Demo, 
           Value_e) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))

df2.2.1 <- rbind(df2.2.t, df2.2.d)

df2.2.1$A_Demo <- factor(df2.2.1$A_Demo,levels = demo_list)

```

```{r, fig.width=12, fig.height=20}
df2.2.1 %>%
  drop_na(A_Demo) %>%
  ggplot(aes(y=pct, x=A_Demo)) + 
  geom_col(aes(fill = Value_e), width = 0.7) + 
  scale_y_continuous(name = "Percentage (%)", label = percent) +
  theme_light() + 
  facet_wrap(~Technology_specific) + 
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 14)) +
  labs(title="Experience") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("A lot", "Some", "None", "Not sure/prefer not to say"), 
                                values = c(
                               "#00A087FF",
                               "#91D1C2FF",
                               "#E64B35FF",
                               "#D5DFE8"),
                    name = "Response") +
  coord_flip()

# experience x age specific stats 
#MP 
#BC
#SMC
#smp
df1 %>%
  drop_na(EXP_MP) %>% # change for each use of AI interested in  
  drop_na(CS_Age_Band) %>%
  filter(CS_Age_Band== 7) %>%
  group_by(EXP_MP) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))


```

```{r}
## tidy up console by removing data frames we no longer need
rm(df2.1,
   df2.1.1,
   df2.1.d,
   df2.1.t,
   df2.2,
   df2.2.1,
   df2.2.d,
   df2.2.t,
   tab)
```

#######################################################################################################################################

Overall ratings of benefit and concern
```{r}
### prep data
## overall levels of benefit 
df3.1 <- df1 %>%
  # remove all labels
  remove_val_labels() %>%
  ## select vars
  select(Respondent_ID, PV25_Weight, 
    tidyselect::vars_select(names(df1), starts_with("BENA_", ignore.case = TRUE)),) %>%
  ## pivot longer
  pivot_longer( # convert to long format 
    names_to = "Question_b",
    values_to = "Value_b",
    BENA_MP:BENA_VR
  ) 

## create new labelling var for specific technologies
df3.1$Technology_specific <- sub('.*_', '', df3.1$Question_b) # extract everything after the underscore into a new variable by replacing everything before the underscore with blank space.



df3.1 <- df3.1 %>%
  mutate("Value_b" = 
  dplyr::recode(Value_b, 
      "1" = "Very",
      "2" = "Somewhat",
      "3" = "Not very",
      "4" = "Not at all",
      "5" = "Don't know/prefer not to say",
      "6" = "Don't know/prefer not to say"))

df3.1$Value_b <- factor(df3.1$Value_b,levels = c(  "Not at all", "Not very", "Don't know/prefer not to say", "Somewhat", "Very")) 


## repeat for concerns
df3.2 <- df1 %>%
  # remove all labels
  remove_val_labels() %>%
  ## select vars
  select(Respondent_ID, PV25_Weight, 
    tidyselect::vars_select(names(df1), starts_with("CONA_", ignore.case = TRUE)),) %>%
  ## pivot longer
  pivot_longer( # convert to long format 
    names_to = "Question_c",
    values_to = "Value_c",
    CONA_MP:CONA_VR
  ) 

## create new labelling var for specific technologies
df3.2$Technology_specific <- sub('.*_', '', df3.2$Question_c) # extract everything after the underscore into a new variable by replacing everything before the underscore with blank space.

df3.2 <- df3.2 %>%
  mutate("Value_c" = 
  dplyr::recode(Value_c, 
      "1" = "Very",
      "2" = "Somewhat",
      "3" = "Not very",
      "4" = "Not at all",
      "5" = "Don't know/prefer not to say",
      "6" = "Don't know/prefer not to say"))

df3.2$Value_c <- factor(df3.2$Value_c,levels = c(  "Not at all", "Not very", "Don't know/prefer not to say", "Somewhat", "Very")) 

df3 <- inner_join(df3.1, df3.2)

df3 <- df3 %>%
  mutate(Technology_specific = 
           recode_technologies(Technology_specific))

rm(df3.1, df3.2)
```

```{r}
## bar plot of benefit levels
df3 %>%
  drop_na(Value_b) %>%
  group_by(Technology_specific, Value_b) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value_b == "Very" | Value_b == "Somewhat" ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Technology_specific, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Technology_specific, -Sum_f), fill = Value_b)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(#panel.grid.major = element_blank(), 
        panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 12)) +
  labs(title="Technology as beneficial",
       subtitle = "'To what extent do you think that the use of this technology will be beneficial?'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Very", "Somewhat", "Don't know/prefer not to say", "Not very", "Not at all" ), 
                                values = c(
                               "#00A087FF",
                               "#91D1C2FF",
                               "#D5DFE8",
                               "#F39B7FFF",
                               "#E64B35FF"
                               ),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip()
```
```{r}
## bar plot of concerns
df3 %>%
  drop_na(Value_c) %>%
  group_by(Technology_specific, Value_c) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value_c == "Very" | Value_c == "Somewhat" ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Technology_specific, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Technology_specific, -Sum_f), fill = Value_c)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(#panel.grid.major = element_blank(), 
        panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 12)) +
  labs(title="Technology as concerning",
       subtitle = "'To what extent are you concerned about the use of...'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Very", "Somewhat", "Don't know/prefer not to say", "Not very", "Not at all" ), 
                                values = c(
                                  "#E64B35FF",
                                  "#F39B7FFF",
                                  "#D5DFE8",
                               "#91D1C2FF",
                               "#00A087FF"
                               ),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip()
```


```{r}
rm(df3)
```

####### Overall - specific benefits and concerns for each technology ############### 
```{r}
## loop through all technologies 
# overall sample specific benefits and concerns 
df5 <- df1 %>%
  select(Respondent_ID, PV25_Weight,
         tidyselect::vars_select(names(df1), starts_with("BENB_", ignore.case = TRUE)),) %>%
  select(!BENB_MP_3) %>%
  select(!BENB_BC_7) %>%
  select(!BENB_PS_6) %>% 
  select(!BENB_WEL_7) %>% 
  select(!BENB_JOB_7) %>% 
  select(!BENB_CAN_6) %>% 
  select(!BENB_LOAN_7) %>% 
  select(!BENB_SMC_4) %>% 
  select(!BENB_SMP_5) %>% 
  select(!BENB_VASS_6) %>% 
  select(!BENB_VAH_7) %>% 
  select(!BENB_RVC_4) %>% 
  select(!BENB_RCA_5) %>% 
  select(!BENB_DC_7) %>% 
  select(!BENB_AW_7) %>% 
  select(!BENB_CCR_5) %>%
  select(!BENB_VR_5) %>%
  pivot_longer( # convert to long format 
    names_to = "Question",
    values_to = "Value",
    BENB_MP01:BENB_VR09)

# broad grouping variable
df5 <- df5 %>%
  mutate(Technology_broad = 
           case_when(
                     str_detect(Question, "_MP") | str_detect(Question,"_BC") | str_detect(Question,"_PS") ~ "Face recognition",
                     str_detect(Question, "_WEL") | str_detect(Question,"_JOB") | str_detect(Question,"_CAN") | str_detect(Question,"_LOAN") ~ "AI risk or eligibility",
                     str_detect(Question, "_SMC") | str_detect(Question,"_SMP") ~ "Social media targeting",
                     str_detect(Question, "_VASS") | str_detect(Question,"_VAH") ~ "Virtual assistants",
                     str_detect(Question, "_RVC") | str_detect(Question,"_RCA") | str_detect(Question,"_DC") | str_detect(Question,"_AW") ~ "Robotics",
                     str_detect(Question, "_CCR") | str_detect(Question,"_VR") ~ "Simulations",
           ))

## create new labelling var for specific technologies
df5$Technology_specific <- sub('.*_', '', df5$Question) # extract everything after the underscore into a new variable 

df5$Technology_specific <- str_sub(df5$Technology_specific, end = -3) # remove last 2 characters from string 

df5 <- df5 %>%
  mutate(Technology_specific = 
           recode_technologies(Technology_specific))
```

```{r}
df5.b <- df1 %>%
  select(Respondent_ID, PV25_Weight,
         tidyselect::vars_select(names(df1), starts_with("CONB_", ignore.case = TRUE)),) %>%
 select(!CONB_MP_7) %>%
  select(!CONB_BC_11) %>%
  select(!CONB_PS_10) %>% 
  select(!CONB_WEL_11) %>% 
  select(!CONB_JOB_11) %>% 
  select(!CONB_CAN_9) %>% 
  select(!CONB_LOAN_12) %>% 
  select(!CONB_SMC_5) %>% 
  select(!CONB_SMP_6) %>% 
  select(!CONB_VASS_7) %>% 
  select(!CONB_VAH_11) %>% 
  select(!CONB_RVC_8) %>% 
  select(!CONB_RCA_9) %>% 
  select(!CONB_DC_11) %>% 
  select(!CONB_AW_10) %>% 
  select(!CONB_CCR_5) %>%
  select(!CONB_VR_6) %>%
  pivot_longer( # convert to long format 
    names_to = "Question",
    values_to = "Value",
    CONB_MP01:CONB_VR12)

# broad grouping variable
df5.b <- df5.b %>%
  mutate(Technology_broad = 
           case_when(
                     str_detect(Question, "_MP") | str_detect(Question,"_BC") | str_detect(Question,"_PS") ~ "Face recognition",
                     str_detect(Question, "_WEL") | str_detect(Question,"_JOB") | str_detect(Question,"_CAN") | str_detect(Question,"_LOAN") ~ "AI risk or eligibility",
                     str_detect(Question, "_SMC") | str_detect(Question,"_SMP") ~ "Social media targeting",
                     str_detect(Question, "_VASS") | str_detect(Question,"_VAH") ~ "Virtual assistants",
                     str_detect(Question, "_RVC") | str_detect(Question,"_RCA") | str_detect(Question,"_DC") | str_detect(Question,"_AW") ~ "Robotics",
                     str_detect(Question, "_CCR") | str_detect(Question,"_VR") ~ "Simulations",
           ))

## create new labelling var for specific technologies
df5.b$Technology_specific <- sub('.*_', '', df5.b$Question) # extract everything after the underscore into a new variable 

df5.b$Technology_specific <- str_sub(df5.b$Technology_specific, end = -3) # remove last 2 characters from string 

df5.b <- df5.b %>%
  mutate(Technology_specific = 
           recode_technologies(Technology_specific))
```

```{r}
## checking specific stats
df1 %>%
  filter(CONA_VASS == 1 | CONA_VASS == 2) %>%
  group_by(CONB_VASS06) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))
```


```{r}
## specific benefits - risk and eligibility
df5 %>%
  filter(Technology_broad == "AI risk or eligibility") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - risk and eligibility
df5.b %>%
  filter(Technology_broad == "AI risk or eligibility") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific benefits - Face recognitio
df5 %>%
  filter(Technology_broad == "Face recognition") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - Face recognitio
df5.b %>%
  filter(Technology_broad == "Face recognition") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct)) 
```

```{r}
## specific benefits - Robotics
df5 %>%
  filter(Technology_broad == "Robotics") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - Robotics
df5.b %>%
  filter(Technology_broad == "Robotics") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific benefits - Virtual assistants
df5 %>%
  filter(Technology_broad == "Virtual assistants") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - Virtual assistants
df5.b %>%
  filter(Technology_broad == "Virtual assistants") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific benefits - Social media targeting
df5 %>%
  filter(Technology_broad == "Social media targeting") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - Social media targeting
df5.b %>%
  filter(Technology_broad == "Social media targeting") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```


```{r}
## specific benefits - Simulations
df5 %>%
  filter(Technology_broad == "Simulations") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

```{r}
## specific concerns - Simulations
df5.b %>%
  filter(Technology_broad == "Simulations") %>%
  drop_na(Value) %>%
  group_by(Technology_specific, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  #### select top three benefits of technology type only #### 
  group_by(Technology_specific) %>%
  arrange(desc(pct))
```

####### Chi-square testing of specific benefits and concerns #######
```{r}
## for this analysis, word documents with chi-square test tables will be produced and saved to your working directory. 
df2 <- df1 %>%
  select(RS_Ethnicity_Combined,
         DR_Socio_Eco,
         RS_Sex,
         CS_Age_Band,
         RS_EducationLevel,
         AWARE_VR,
         PV25_Weight,
         tidyselect::vars_select(names(df1), starts_with("BENB_", ignore.case = TRUE)),
         tidyselect::vars_select(names(df1), starts_with("CONB_", ignore.case = TRUE))) %>%
  mutate(RS_Ethnicity_Combined = as.factor(RS_Ethnicity_Combined)) %>%
  mutate(DR_Socio_Eco = as.factor(DR_Socio_Eco)) %>%
  mutate(RS_Sex = as.factor(RS_Sex)) %>%
  mutate(CS_Age_Band = as.factor(CS_Age_Band)) %>%
  mutate(RS_EducationLevel = as.factor(RS_EducationLevel)) %>%
  mutate(AWARE_VR = as.factor(AWARE_VR))

df2 <- df2 %>%
  select(!BENB_MP_3) %>%
  select(!BENB_BC_7) %>%
  select(!BENB_PS_6) %>% 
  select(!BENB_WEL_7) %>% 
  select(!BENB_JOB_7) %>% 
  select(!BENB_CAN_6) %>% 
  select(!BENB_LOAN_7) %>% 
  select(!BENB_SMC_4) %>% 
  select(!BENB_SMP_5) %>% 
  select(!BENB_VASS_6) %>% 
  select(!BENB_VAH_7) %>% 
  select(!BENB_RVC_4) %>% 
  select(!BENB_RCA_5) %>% 
  select(!BENB_DC_7) %>% 
  select(!BENB_AW_7) %>% 
  select(!BENB_CCR_5) %>%
  select(!BENB_VR_5)

df2 <- df2 %>%
  select(!CONB_MP_7) %>%
  select(!CONB_BC_11) %>%
  select(!CONB_PS_10) %>% 
  select(!CONB_WEL_11) %>% 
  select(!CONB_JOB_11) %>% 
  select(!CONB_CAN_9) %>% 
  select(!CONB_LOAN_12) %>% 
  select(!CONB_SMC_5) %>% 
  select(!CONB_SMP_6) %>% 
  select(!CONB_VASS_7) %>% 
  select(!CONB_VAH_11) %>% 
  select(!CONB_RVC_8) %>% 
  select(!CONB_RCA_9) %>% 
  select(!CONB_DC_11) %>% 
  select(!CONB_AW_10) %>% 
  select(!CONB_CCR_5) %>%
  select(!CONB_VR_6)

### column numbers to reference in code chunk below! 

colnames(df2)

```

```{r}
## SES 
for (col in names(df2)[8:18]) { ## change number in this line for columns you are interested in (refer to codebook)
  sink(file = paste(col, "SES", sep = "_", ".doc"))
  ct_ <- crosstab(df2[[col]], 
                  df2[["DR_Socio_Eco"]], 
                  weight = df2[["PV25_Weight"]], 
                  format = "SPSS", 
                  prop.c = TRUE, 
                  plot = FALSE,
                  chisq = TRUE, fisher = TRUE)
  ct_[["RowData"]] <- col
  ct_[["ColData"]] <- "DR_Socio_Eco"
  print(ct_)
  sink()
}


## eth 
 for (col in names(df2)[8:18]) { ## change number in this line for columns you are interested in (refer to codebook)
  sink(file = paste(col, "ETH", sep = "_", ".doc"))
  ct_ <- crosstab(df2[[col]], 
                  df2[["RS_Ethnicity_Combined"]], 
                  weight = df2[["PV25_Weight"]], 
                  format = "SPSS", 
                  prop.c = TRUE, 
                  plot = FALSE,
                  chisq = TRUE, fisher = TRUE)
  ct_[["RowData"]] <- col
  ct_[["ColData"]] <- "RS_Ethnicity_Combined"
  print(ct_)
  sink()
}


## sex 
for (col in names(df2)[8:18]) { ## change number in this line for columns you are interested in (refer to codebook)
  sink(file = paste(col, "SEX", sep = "_", ".doc"))
  ct_ <- crosstab(df2[[col]], 
                  df2[["RS_Sex"]], 
                  weight = df2[["PV25_Weight"]], 
                  format = "SPSS", 
                  prop.c = TRUE, 
                  plot = FALSE,
                  chisq = TRUE, fisher = TRUE)
  ct_[["RowData"]] <- col
  ct_[["ColData"]] <- "RS_Sex"
  print(ct_)
  sink()
}

## age 
for (col in names(df2)[8:18]) { ## change number in this line for columns you are interested in (refer to codebook)
  sink(file = paste(col, "AGE", sep = "_", ".doc"))
  ct_ <- crosstab(df2[[col]], 
                  df2[["CS_Age_Band"]], 
                  weight = df2[["PV25_Weight"]], 
                  format = "SPSS", 
                  prop.c = TRUE, 
                  plot = FALSE,
                  chisq = TRUE, fisher = TRUE)
  ct_[["RowData"]] <- col
  ct_[["ColData"]] <- "CS_Age_Band"
  print(ct_)
  sink()
}


## education 
for (col in names(df2)[8:18]) { ## change number in this line for columns you are interested in (refer to codebook)
  sink(file = paste(col, "EDU", sep = "_", ".doc"))
  ct_ <- crosstab(df2[[col]], 
                  df2[["RS_EducationLevel"]], 
                  weight = df2[["PV25_Weight"]], 
                  format = "SPSS", 
                  prop.c = TRUE, 
                  plot = FALSE,
                  chisq = TRUE, fisher = TRUE)
  ct_[["RowData"]] <- col
  ct_[["ColData"]] <- "RS_EducationLevel"
  print(ct_)
  sink()
}
```

```{r}
rm(df2, df5, df5.b)
```


######### Governance, comfort, explainability ###############
```{r}
# Explainability vs accuracy 
df6 <- df1 %>%
  select(Respondent_ID, PV25_Weight, CS_Age_Band, 
         EXPLAINB)

# prep data
df6 <- df6 %>%
  mutate("EXPLAINB" = 
  dplyr::recode(EXPLAINB, 
      "1" = "Accuracy is more important than providing an explanation",
      "2" = "Sometimes explanation should be given, even if that makes the AI decision less accurate",
      "3" = "An explanation should always be given, even if that makes all AI decisions less accurate",
      "4" = "Humans, not computers, should always make the decisions and be able to explain them to the people affected",
      "5" = "Don't know/prefer not to say",
      "6" = "Don't know/prefer not to say"))

## create total summary table
df6.t <- df6 %>%
  drop_na(EXPLAINB) %>%
  group_by(EXPLAINB) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))

df6.t$CS_Age_Band <- "Total"
## create a summary table by age
df6.a <- df6 %>%
  drop_na(EXPLAINB) %>%
  group_by(CS_Age_Band, 
           EXPLAINB) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n))

## merge tables
df6.all <- rbind(df6.t, df6.a)

## relevel the data - explain b
df6.all$EXPLAINB <- factor(df6.all$EXPLAINB,levels = c( "Don't know/prefer not to say",  
                                                        "Humans, not computers, should always make the decisions and be able to explain them to the people affected", 
                                                        "An explanation should always be given, even if that makes all AI decisions less accurate",
                                                        "Sometimes explanation should be given, even if that makes the AI decision less accurate",
                                                        "Accuracy is more important than providing an explanation"))
## relvel the data - demographics 
df6.all <- df6.all %>%
  mutate(CS_Age_Band =
           recode(CS_Age_Band, 
      "1" = "18-24 yrs",
      "2" = "25-34 yrs",
      "3" = "35-44 yrs",
      "4" = "45-54 yrs",
      "5" = "55-64 yrs", 
      "6" = "65-74 yrs",
      "7" = "75+ yrs")) 

df6.all$CS_Age_Band <- factor(df6.all$CS_Age_Band,levels = c("75+ yrs",
                                                             "65-74 yrs",
                                                             "55-64 yrs",
                                                             "45-54 yrs",
                                                             "35-44 yrs",
                                                             "25-34 yrs",
                                                             "18-24 yrs",
                                                             "Total")) 

## overall plot 
df6.all %>%
  ggplot(aes(y=pct, x=CS_Age_Band, fill = EXPLAINB)) + 
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=8), legend.title =  element_text(size = 10)) +
  labs(title="Explainability vs accuracy",
       subtitle = "'Overall, which statement do you feel best reflects your personal opinion?'") + 
  theme(plot.title = element_text(size = 14)) +
 # theme(axis.line.x = element_line(arrow = grid::arrow(length = unit(0.3, "cm"), 
  #                                                     ends = "both"))) +
  scale_fill_manual(breaks = c("Accuracy is more important than providing an explanation", 
                               "Sometimes explanation should be given, even if that makes the AI decision less accurate", 
                               "An explanation should always be given, even if that makes all AI decisions less accurate", 
                               "Humans, not computers, should always make the decisions and be able to explain them to the people affected", 
                               "Don't know/prefer not to say"), 
                                values = c(
                               "#FF6A58",
                               "#FFD2CD",
                               "#D5DFE8",
                               "#7A99B6",
                               "grey"),
                    name = "",
                    labels = function(x) str_wrap(x, width = 20)) +
  xlab("") + 
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip() 
```
```{r}
rm(df6, df6.a, df6.all, df6.t)
```


Who should govern AI? 
```{r}
df7 <- df1 %>%
  select(Respondent_ID, PV25_Weight, CS_Age_Band,
         GOVERNANCE01:GOVERNANCE12
         ) %>%
  pivot_longer(
    names_to = "Question",
    values_to = "Value",
    GOVERNANCE01:GOVERNANCE12
  )

df7 <- df7 %>%
  mutate("Question" = 
  dplyr::recode(Question, 
      "GOVERNANCE01" = "The companies developing the AI technology",
      "GOVERNANCE02" = "Scientists and researchers working in universities or research institutions",
      "GOVERNANCE03" = "The Government",
      "GOVERNANCE04" = "An independent regulator",
      "GOVERNANCE05" = "International standards bodies",
      "GOVERNANCE06" = "An independent oversight committee with citizen involvement",
      "GOVERNANCE07" = "The people using the AI (e.g. companies, public services)",
      "GOVERNANCE08" = "Other (please specify)",
      "GOVERNANCE09" = "No one should be responsible",
      "GOVERNANCE10" = "Don't know/prefer not to say",
      "GOVERNANCE11" = "Don't know/prefer not to say",
      "GOVERNANCE12" = "All of the above")) 

# overall preference
df7 %>%
  drop_na(Value) %>%
  group_by(Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  filter(pct > 0.004) %>%
  ggplot(aes(y=pct, x=reorder(Question, -pct))) + 
  geom_col(fill = "#FF6A58") + 
    scale_y_continuous(name = "", label = percent) +
    geom_text(aes(label = scales::percent(pct, accuracy = 1)), vjust = -0.5) +
    theme_minimal() + 
    theme(axis.text.x = element_text(angle=45, hjust=1)) + 
    coord_cartesian(ylim=c(0, 1))   +
    ggtitle("AI governance",
            subtitle = "Who do you think should be most responsible for ensuring AI is used safely? You can only select up to two options.") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) +
    xlab(" ") 
```



```{r}
# governance by age
df7 <- df7 %>%
  mutate(CS_Age_Band =
           recode(CS_Age_Band, 
      "1" = "18-24 yrs",
      "2" = "25-34 yrs",
      "3" = "35-44 yrs",
      "4" = "45-54 yrs",
      "5" = "55-64 yrs", 
      "6" = "65-74 yrs",
      "7" = "75+ yrs")) 

df7 %>%
  drop_na(Value) %>%
  group_by(CS_Age_Band, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  filter(Question != "Other (please specify)") %>%
  ggplot(aes(y=pct, x=reorder(Question, +pct))) + 
  geom_col(fill = "#FF6A58") + 
    scale_y_continuous(name = "", label = percent) +
    geom_text(aes(label = scales::percent(pct, accuracy = 1)), position = position_dodge(width = .9), size = 3) +
    theme_minimal() + 
    theme(
        axis.text.x = element_text(angle=45, hjust=1)) + 
    coord_cartesian(ylim=c(0, 1))   +
    facet_grid(~CS_Age_Band) + 
    ggtitle("AI governance",
            subtitle = "Who do you think should be most responsible for ensuring AI is used safely? You can only select up to two options.") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) +
    xlab(" ") +
  coord_flip()

df7 %>%
  drop_na(Value) %>%
  group_by(CS_Age_Band, Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Question == "The companies developing the AI technology") %>%
  group_by(CS_Age_Band) %>%
  mutate(total = sum(n))
```
```{r}
rm(df7)
```

comfort with AI 
```{r, echo=FALSE, warning=FALSE}
df8 <- df1 %>%
  select(Respondent_ID, PV25_Weight, CS_Age_Band,
         MORECOMF01:MORECOMF12
         ) %>%
  pivot_longer(
    names_to = "Question",
    values_to = "Value",
    MORECOMF01:MORECOMF12
  )

df8 <- df8 %>%
  mutate("Question" = 
  dplyr::recode(Question, 
      "MORECOMF01" = "Explanations of how AI technologies work and make decisions",
      "MORECOMF02" = "More human involvement and control in AI decisions",
      "MORECOMF03" = "Clear procedures in place for appealing to a human specialist against a decision made by AI",
      "MORECOMF04" = "The AI has been deemed acceptable by a government regulator",
      "MORECOMF05" = "Laws and regulations",
      "MORECOMF06" = "People’s personal information is kept safe and secure",
      "MORECOMF07" = "Regular evaluation to ensure no discrimination against particular groups of people",
      "MORECOMF08" = "Something else (please specify)",
      "MORECOMF09" = "None of these",
      "MORECOMF10" = "Nothing",
      "MORECOMF11" = "Don't know/prefer not to say",
      "MORECOMF12" = "Don't know/prefer not to say")) 

df8 %>%
  drop_na(Value) %>%
  group_by(Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  filter(Value == 1) %>%
  ggplot(aes(y=pct, x=reorder(Question, -pct))) + 
  geom_col(fill = "#FF6A58") + 
    scale_y_continuous(name = "", label = percent) +
    geom_text(aes(label = scales::percent(pct, accuracy = 1)), vjust = -0.5) +
    theme_minimal() + 
    theme(
        axis.text.x = element_text(angle=45, hjust=1)) + 
    coord_cartesian(ylim=c(0, 1))   +
    ggtitle("Comfort with AI",
            subtitle = "'Which of the following, if any, would make you more comfortable with AI technologies being used? Please select all that apply'") +
  scale_x_discrete(labels = function(x) str_wrap(x, width = 25)) +
    xlab(" ") 
```
```{r}
rm(df8)
```


Who should develop AI? Cancer risk detection
```{r}
dev.df <- df1 %>%
  select(Respondent_ID, PV25_Weight, 
         RS_Ethnicity_Combined,
         CS_Age_Band,
         RS_Sex,
         RS_EducationLevel,
         DR_Socio_Eco,
         DEV_CAN_1:DEV_CAN_4
  ) %>%
  remove_val_labels() %>%
  pivot_longer(names_to = "Question",
               values_to = "Value",
                DEV_CAN_1:DEV_CAN_4)

dev.df <- dev.df %>%
  mutate("Value" = 
  dplyr::recode(Value, 
      "1" = "Very",
      "2" = "Somewhat",
      "3" = "Not very",
      "4" = "Not at all",
      "5" = "Don't know/prefer not to say",
      "6" = "Don't know/prefer not to say"))

dev.df <- dev.df %>%
  mutate("Question" = 
  dplyr::recode(Question, 
      "DEV_CAN_1" = "Private companies",
      "DEV_CAN_2" = "Not fo profit organisations",
      "DEV_CAN_3" = "A governmental body",
      "DEV_CAN_4" = "Universities/academic researchers"))

dev.df$Value <- factor(dev.df$Value,levels = c("Don't know/prefer not to say",
                                               "Not at all",
                                               "Not very",
                                               "Somewhat",
                                               "Very"
                                               ))

dev.df %>%
  drop_na(Value) %>%
  group_by(Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value == "Very" | Value == "Somewhat" ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Question, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Question, -Sum_f), fill = Value)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), 
                        legend.position = "top", legend.text = element_text(size=12), legend.title =  element_text(size = 12)) +
  labs(title="Cancer risk technology development",
       subtitle = "'How concerned do you feel, if at all, about each of them producing AI technologies that predict the risk of developing cancer?'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Very", "Somewhat", "Not very", "Not at all", "Don't know/prefer not to say"), 
                                values = c(
                                  "#E64B35FF",
                                  "#F39B7FFF",
                                  "#91D1C2FF",
                               "#00A087FF",
                               "#D5DFE8"),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip()
```

Who should develop AI? Welfar eligiblity
```{r}
dev.df.1 <- df1 %>%
  select(Respondent_ID, PV25_Weight, 
         RS_Ethnicity_Combined,
         CS_Age_Band,
         RS_Sex,
         RS_EducationLevel,
         DR_Socio_Eco,
         DEV_WELFARE_1:DEV_WELFARE_4
  ) %>%
  remove_val_labels() %>%
  pivot_longer(names_to = "Question",
               values_to = "Value",
                DEV_WELFARE_1:DEV_WELFARE_4)

dev.df.1 <- dev.df.1 %>%
  mutate("Value" = 
  dplyr::recode(Value, 
      "1" = "Very",
      "2" = "Somewhat",
      "3" = "Not very",
      "4" = "Not at all",
      "5" = "Don't know/prefer not to say",
      "6" = "Don't know/prefer not to say"))

dev.df.1 <- dev.df.1 %>%
  mutate("Question" = 
  dplyr::recode(Question, 
      "DEV_WELFARE_1" = "Private companies",
      "DEV_WELFARE_2" = "Not fo profit organisations",
      "DEV_WELFARE_3" = "A governmental body",
      "DEV_WELFARE_4" = "Universities/academic researchers"))

dev.df.1$Value <- factor(dev.df.1$Value,levels = c("Don't know/prefer not to say",
                                               "Not at all",
                                               "Not very",
                                               
                                               "Somewhat",
                                               "Very"
                                               ))
```

```{r}
dev.df.1 %>%
  drop_na(Value) %>%
  group_by(Question, Value) %>%
  tally(wt = PV25_Weight) %>%
  mutate(pct = n/sum(n)) %>%
  mutate(Value_factor = 
           case_when(Value == "Very" | Value == "Somewhat" ~ "High", 
                     TRUE ~ "Other")) %>%
  group_by(Question, Value_factor) %>%
  mutate(Sum_f = sum(pct)) %>%
  #### ggplot #####
  ggplot(aes(y=pct, x=reorder(Question, -Sum_f), fill = Value)) +
  geom_col(width = 0.7) + 
  scale_y_continuous(name = "", label = percent) +
  theme_minimal() + 
  theme(#panel.grid.major = element_blank(), 
        panel.grid.minor = element_blank(),
                        legend.position = "top", legend.text = element_text(size=8), legend.title =  element_text(size = 12)) +
  labs(title="Welfare eligibility technology development",
       subtitle = "'How concerned do you feel, if at all, about each of them producing new computer technologies for assessing eligibility for welfare benefits?'") + 
  theme(plot.title = element_text(size = 14)) +
  scale_fill_manual(breaks = c("Very", "Somewhat", "Not very", "Not at all", "Don't know/prefer not to say"), 
                                values = c(
                                  "#E64B35FF",
                                  "#F39B7FFF",
                                  "#91D1C2FF",
                               "#00A087FF",
                               "#D5DFE8"),
                    name = "Response") +
  xlab("") +
  geom_text(
    aes(label = scales::percent(pct, accuracy = 1)), position = position_stack(vjust = .5), size = 3) +
  coord_flip(clip= "off")
```