---
title: "Exercises on plots"
output: html_notebook
---

**DISCLAIMER**: the graphs in this series of exercises are pretty ugly!


# Setting the environment

First, you need to install the packages seen in class. **AND DON'T FORGET TO SET YOUR WORKING DIRECTORY!**

```{r install, message = FALSE, warning = FALSE}
install.packages("plotly")
install.packages("gganimate")
install.packages("ggrepel")
install.packages("cowplot")
install.packages("gifski")
install.packages("png")
```

Then, you need to activate them!

```{r activate, message = FALSE, warning = FALSE}
library(tidyverse)
library(plotly)
library(gganimate)
library(ggrepel)
library(cowplot)
library(gifski)
library(png)
```

Finally, you need set **YOUR** working directory and load the datasets.  
You can do that directly in the files pane in RStudio (e.g., click on the blue wheel!).

```{r load}
# setwd("~/Documents/IT/COURS/2021/R4DS/S3_plots")  # That works only for me!!!
load("movies.RData")
load("songs.RData")
```


# The exercises

## Songs dataset

1. Plot the histogram of the **popularity** variable. **Hint**: it's geom_histogram().


2. Create a pivot table that lists the top artists with more than 19 songs in the dataset and plot their number of songs with bars. **Hint**: it's geom_col().
           
           
3. Create a scatter plot (with points) where the *x*-axis is the length of the song and the *y*-axis is the popularity. You can set the point size to 0.5. Add a geom_smooth.

           
4. 
a) Filter the data to keep only songs from Lady Gaga, Drake and Kanye West.   
b) Use geom_point to show the relationship between *speechiness* and *energy* and use color for artists.
c) Use geom_smooth to detect the local patterns. 

           
5. **Impact of energy & valence on top popular songs**.
a) Filter songs with popularity above 60.
b) On top of that, create a graph with points such that *x* = energy and *y* = valence.
c) Add a layer with only the songs with popularity larger than 90, shown in red.
d) Save this plot in a variable and enhance it with ggplotly().



6. **Comparing artists on some key metrics**.
a) Like in Q4, filter artists to keep data for Lady Gaga, Drake, Kanye West and Eminem.
b) Create a pivot table over them that computes average popularity, average danceability and average speechiness.
c) The output is not tidy! Transform it into tidy format (create an indicator variable that will encompass the averages). 
d) Plot the values via geom_col() and facet_grid(). You can pick a color for each artist. 

           
           
## Movies dataset

1. Plot the histogram of the **imdb_score** variable. 

2. Create a pivot table that lists the top directors with more than 15 films in the dataset and plot their number of movies with bars. 

3. Create a scatter plot (with points) where the *x*-axis is the length of the movie and the *y*-axis is the imdb_score You can set the point size to 0.5. Add a geom_smooth.

4. Create a pivot table which computes, for each year, the average popularity of the movies (of that year). Create an animation that displays the corresponding figures through time with bars. **Hint**: use the layers geom_col(), transition_time() - this one is new - and shadow_mark().

5. 
a) Filter the data to keep only films from Steven Spielberg, Woody Allen and Clint Eastwood.   
b) Use geom_point to show the relationship between *budget* and *imdb_score* and use color for director.
c) Use geom_smooth to detect the local patterns. 

6. **Impact of nb_voters & earnings/budget on best rated films**.
a) Filter movies with imdb_score above 5.5.
b) On top of that, create a graph with points such that *x* = nb_voters and *y* = earnings/budget.
c) Add a layer with only the films with imdb_score larger than 8, shown in red (add a label=director in the aes()).
d) Change both axes into log10 scales.
e) Save this plot in a variable and enhance it with ggplotly().

7. **Comparing directors on some key metrics**.
a) Like in Q5, filter directors to keep data for Steven Spielberg, Woody Allen, Clint Eastwood, Martin Scorsese and Ridley Scott
b) Create a pivot table over them that computes average earnings/budget, average duration and average imdb_score.
c) The output is not tidy! Transform it into tidy format (create an indicator variable that will encompass the averages) 
d) Plot the values via geom_col() and facet_grid(). You can use an angle for the x-axis to ease readability.