notebook.qmd

---
title: "Reproduction of figures from Diamant et al. 2024"
author: "Jackson Vanfleet-Brown"
format: html
echo: false
---

```{r message=FALSE}

library(tidyverse)
library(tidymodels)
library(targets)

tar_load(c("study_data", "combined_predictions"))
```

## Figure 4

Reproduction shown below in @fig-fig4

```{r}
#| label: fig-fig4

ggplot(study_data)+
  geom_histogram(aes(x=whistle_num, fill = label),
                 binwidth = 10,
                 position=position_dodge())
```

## Figure 5

Reproduction shown below in @fig-fig5

```{r results="hide"}
#| label: fig-fig5
#| fig-cap: "Histograms of whistle features, with and without vessel."
#| fig-subcap: 
#|     - "Whistle length"
#|     - "Whistle overlap"
#|     - "Number of Clusters"
#|     - "Harmony Rate"

vars_wanted <- c("whistle_length", "harmony_num", "whistle_overlap", "cluster_num")

binwidths <- c(0.05, 1, 1, 1)

make_plot <- function(x, y) {
  ggplot(x) +
    geom_histogram(aes(x= value, fill = label),
                   position = position_dodge(),
                   binwidth = binwidths[grep(y, vars_wanted)]
    )
}

study_data %>%
  pivot_longer(cols=1:9, names_to = "var") %>%
  nest(data=-var) %>%
  filter(var %in% vars_wanted) %>%
  mutate(plot = map2(data, var, ~ make_plot(.x, .y))) %>%
  pull(plot)


```

## Prediction Results

Reproduction of Figure 6 shown in @fig-fig6 below.

```{r}
#| label: fig-fig6
#| fig-cap: "Classification results"
#| 
metrics_wanted <- c("accuracy", "sens")

results <- combined_predictions %>%
  nest(data = -svm) %>%
  mutate(cf = map(data, ~ conf_mat(.x, truth, est))) %>%
  mutate(summary = map(cf, ~ summary(.x))) %>%
  mutate(metrics = map(summary, ~ filter(.x, .metric %in% metrics_wanted))) %>%
  pull(metrics, name = "svm") %>%
  list_rbind(names_to = "svm")

ggplot(results) +
  geom_bar(mapping = aes(x=svm, y=.estimate, fill = .metric),
           position = position_dodge(),
           stat = "identity")

```