homework-solutions.rmd

---
title: "CEU DV3: Example homework solutions"
author: "Gergely Daroczi"
date: "January 24, 2024"
output: html_document
---

```{r suppress, echo = FALSE}
knitr::opts_chunk$set(echo = FALSE, fig.height = 10, fig.width = 15)
```

## 0. Load the `nycflights13` package and check what datasets are bundled.


```{r}
library(nycflights13)
knitr::kable(data(package = 'nycflights13')$results)
```

## 1. Visualize the distribution of arrival delays per origin airport!

```{r}
library(ggplot2)
ggplot(flights, aes(origin, arr_delay)) + geom_boxplot()
```

## 2. Visualize the distribution of arrival delays per destination airport! Note that the x axis labels need to be rotated, and spend some time cleaning up the axis and plot titles! Might need to tweak `fig.height` and `fig.width` params of the code chunk in Rmd!

```{r}
ggplot(flights, aes(dest, arr_delay)) + geom_boxplot() +
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)) +
    xlab('') + ylab('Arrival delays (minutes)') +
    ggtitle('Flights deparing from NY in 2013')
```

## 3. Compute and visualize the average arrival delay per destination! Make sure to handle `NA`s and order the barplot by the average delay.

```{r}
library(data.table)
dt <- data.table(flights)[, .(delay = mean(arr_delay, na.rm = TRUE)), by = dest][order(delay)]
dt[, dest := factor(dest, levels = dest)]
ggplot(
    dt,
    aes(dest, delay)) + geom_col() +
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)) +
    xlab('') + ylab('Average arrival delays (minutes)') +
    ggtitle('Flights deparing from NY in 2013')
```

Using `reorder` without an explicit `factor` call:

```{r}
dt <- data.table(flights)[, .(delay = mean(arr_delay, na.rm = TRUE)), by = dest][order(delay)]
ggplot(
    dt,
    aes(reorder(dest, delay), delay)) + geom_col() +
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)) +
    xlab('') + ylab('Average arrival delays (minutes)') +
    ggtitle('Flights deparing from NY in 2013')
```

Note that this is not a `ggplot2` trick, but using `reorder` from base R to set the factor levels on the fly instead of making it through the dataset.

## 4. Redo the above plot by showing the actual name of the destination airport instead of it's FAA id!

```{r}
dt <- merge(
    data.table(flights),
    data.table(airports)[, .(dest = faa, name)])
dt <- dt[, .(delay = mean(arr_delay, na.rm = TRUE)), by = name][order(delay)]
dt[, name := factor(name, levels = name)]
ggplot(
    dt,
    aes(name, delay)) + geom_col() +
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)) +
    xlab('') + ylab('Average arrival delays (minutes)') +
    ggtitle('Flights deparing from NY in 2013')
```

## 5. Color the bars by the timezone of the airport! Make sure to render the legend on the top, using a single line.

```{r}
dt <- merge(
    data.table(flights),
    data.table(airports)[, .(dest = faa, name, tzone)])
dt <- dt[, .(delay = mean(arr_delay, na.rm = TRUE)), by = .(name, tzone)][order(delay)]
dt[, name := factor(name, levels = name)]
ggplot(
    dt,
    aes(name, delay, fill = tzone)) + geom_col() +
    theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1)) +
    xlab('') + ylab('Average arrival delays (minutes)') +
    theme(legend.position = 'top') +
    guides(fill = guide_legend(title = 'Timezone', nrow = 1)) +
    ggtitle('Flights deparing from NY in 2013')
```

## 6. Geocode the destination airports, then visualize those on a worldmap with the point sizes relative to the number of flights to there from NY!


```{r}
dt <- data.table(flights)[, .N, by = dest]
dt[, address := paste(dest, 'airport')]
dt <- tidygeocoder::geocode(dt, 'address')
saveRDS(dt, '/tmp/dt.rda')

world <- map_data('world')
ggplot() +
    geom_map(data = world, map = world, aes(long, lat, map_id = region)) +
    geom_point(data = dt, aes(long, lat, size = N), color = 'orange') +
    coord_fixed(1.3) +
    theme_void()
```

## 7. Compute the average departure and arrival delays, also the average air time and distance per destination, then pass the scaled dataframe through MDS to visualize the similarities/dissimilarities of the destination airports!

```{r}
dt <- data.table(flights)[, .(
    dep_delay = mean(dep_delay, na.rm = TRUE),
    arr_delay = mean(arr_delay, na.rm = TRUE),
    air_time = mean(air_time, na.rm = TRUE),
    distance = mean(distance, na.rm = TRUE)
), by = dest]
mds <- as.data.frame(cmdscale(dist(scale(dt[, 2:5]))))
mds$faa <- dt$dest
library(ggrepel)
ggplot(mds, aes(V1, V2, label = faa)) + geom_text_repel() + theme_void()
```