---
title: Midlevel analysis ($\chi^2$)
author: "Marcin"
editor_options: 
  chunk_output_type: console
output: pdf_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, tidy = FALSE, size = 'footnotesize', fig.align = 'center', out.width = '0.7\\textwidth', cache = FALSE)
```

```{r echo=FALSE, message=FALSE}
library('tidyverse')
theme_set(theme_bw())
```

```{r}
data <- read.csv('data_matrix_items.csv')
data$dim <- as.factor(str_replace(data$dim, '_1', ''))
```

Number of stimuli per subject and category:
```{r}
all(xtabs(~ dim + id, data) == 12)
```

Number of subjects:
```{r}
length(unique(data$id))
```

Number of unique stimuli:
```{r}
length(unique(data$item))
```

The above should be 48 - is there a duplicate?
```{r}
data %>%
  group_by(id, item) %>%
  summarize(reps_per_subj = n()) %>%
  filter(reps_per_subj > 1) %>%
  group_by(item, reps_per_subj) %>%
  count()
```

Remove the second occurrence of the duplicate stimulus:
```{r}
data <- data %>%
  group_by(id, item) %>%
  arrange(X) %>%
  slice(1)
```

# Overall plot

```{r}
data %>%
  pivot_longer(!c('id', 'dim', 'item', 'X')) %>%
  group_by(dim, name) %>%
  summarize(perc_ones = 100 * sum(value) / n()) %>%
  ggplot(aes(x = dim, y = perc_ones)) +
  geom_col() +
  facet_wrap(~ name, ncol = 5) +
  theme(axis.text.x = element_text(angle = 45, vjust = 1, hjust= 1))
```


# $\chi^2$ tests

Function to run multiple comparisons and correct p-values:

```{r}
chisq_multicomp <- function(df) {
  df <- as.data.frame(df)
  res <- matrix(NA, nrow(df) - 1, 3)
  for (i in 2:nrow(df)) {
    chisq <- chisq.test(df[c(1, i), -1])
    res[i-1, 1] <- chisq$statistic
    res[i-1, 2] <- chisq$p.value
  }
  res[,3] <- res[,2] * (nrow(df) - 1)
  res[res[,3] > 1, 3] <- 1
  rownames(res) <- df[-1, 1]
  colnames(res) <- c('chisq', 'p_value', 'p_value_cor')
  return(res)
}
```

# Common ground

```{r}
data$dim <- relevel(data$dim, which(levels(data$dim) == 'dim10_low'))
common_ground <- group_by(data, dim) %>%
  summarize(yes = sum(working.towards.common.ground),
            no = n() - yes)
common_ground
ggplot(common_ground, aes(x = dim, y = yes / (yes + no))) + geom_col()
chisq_multicomp(common_ground)
```

# Explanation

```{r}
data$dim <- relevel(data$dim, which(levels(data$dim) == 'dim10_high'))
explanation <- group_by(data, dim) %>%
  summarize(yes = sum(explanation),
            no = n() - yes)
explanation
ggplot(explanation, aes(x = dim, y = yes / (yes + no))) + geom_col()
chisq_multicomp(explanation)
```

# Self-directed

```{r}
data$dim <- relevel(data$dim, which(levels(data$dim) == 'dim05_low'))
self_dir <- group_by(data, dim) %>%
  summarize(yes = sum(self.directed),
            no = n() - yes)
self_dir
ggplot(self_dir, aes(x = dim, y = yes / (yes + no))) + geom_col()
chisq_multicomp(self_dir)
```

# Other-directed

```{r}
data$dim <- relevel(data$dim, which(levels(data$dim) == 'dim05_high'))
other_dir <- group_by(data, dim) %>%
  summarize(yes = sum(other.directed),
            no = n() - yes)
other_dir
ggplot(other_dir, aes(x = dim, y = yes / (yes + no))) + geom_col()
chisq_multicomp(other_dir)
```