rnotebook_ggplot2.Rmd

---
title: "Plots with ggplot2 are better plots!"
output: html_notebook
---

# Recomended packages
```{r}
install.packages("reshape2", "cowplot", "ggrepel")
```

# Example dataset: Temperature in Barcelona (1780-2017)
## Data
We are going to use a dataset from [Servei Meteorològic de Catalunya](http://www.meteo.cat/wpweb/climatologia/serveis-i-dades-climatiques/serie-climatica-historica-de-barcelona/) describing the mean temperature per month from 1780 to 2017.^[Prohom M, Barriendos, Aguilar E, Ripoll R (2012): Recuperación y análisis de la serie de temperatura diaria de Barcelona, 1780-2011. Cambio Climático. Extremos e Impactos, Asociación Española de Climatología, Serie A, Vol. 8, 207–217]. 

The file that we downloaded has 13 columns: the first one represents the year and the 2nd to the 13th the month. As you can observe, this is a data.frame in wide format, so we will have to convert it to long format. Temperature values are in ºC.

```{r load-data}
meteo <- read.delim("data/Barcelona_TM_m_1780_2017.txt", header=FALSE)
head(meteo)

## Add column names
colnames(meteo) <- c("year", "January", "February", "March", "April", "May", "June",
                     "July", "August", "September", "October", "November", "December")

head(meteo)

## Convert to long format
meteo.long <- reshape2::melt(meteo,
                             id.vars=1,
                             measure.vars=2:ncol(meteo),
                             variable.name="month",
                             value.name="temperature")
head(meteo.long)

## Include a column with season
seasons <- data.frame("month"=unique(meteo.long$month),
                      "season"=c(rep("Winter", 2),
                                 rep("Spring", 3),
                                 rep("Summer", 3),
                                 rep("Fall", 3),
                                 "Winter")
                      )
meteo.long <- merge(meteo.long, seasons)

## Remove NAs (corresponding to Sept-Dec 2017)
meteo.long <- meteo.long[!is.na(meteo.long$temperature),]

head(meteo.long)

```

## Average temperature per month

```{r boxplot}
library(cowplot)

ave <- ggplot(meteo.long, aes(month, temperature)) +
        geom_boxplot()

ave

# Add some more parameters to make it look nicer
ave <- ggplot(meteo.long, aes(month, temperature, color=season)) +
        geom_boxplot(lwd=1) +
        theme(legend.position = "top",
              axis.text.x = element_text(angle=30, vjust=1, hjust=1))
ave

## Change colors
scale_season <- c("Fall"="sienna4",
                  "Spring"="springgreen4",
                  "Summer"="yellow3",
                  "Winter"="navy")

ave <- ave + 
        scale_color_manual(values=scale_season)
ave

## Add title and change labs
ave <- ave + 
        ggtitle("Average Temperature in Barcelona") +
        xlab("Month") +
        ylab("Temperature (ºC)")
ave
```


```{r violin-plot}
ggplot(meteo.long, aes(month, temperature)) +
  geom_violin(aes(color=season), lwd=1) +
  theme(legend.position = "top",
        axis.text.x = element_text(angle=30, vjust=1, hjust=1)) +
  scale_color_manual(values=c(scale_season)) +
  ggtitle("Average Temperature in Barcelona") +
        xlab("Month") +
        ylab("Temperature (ºC)")

## Maybe fill the violin instead of border?
ggplot(meteo.long, aes(month, temperature)) +
  geom_violin(aes(fill=season), lwd=1, alpha=0.7) +
  theme(legend.position = "top",
        axis.text.x = element_text(angle=30, vjust=1, hjust=1)) +
  scale_fill_manual(values=scale_season,
                    name="Season") +
  ggtitle("Average Temperature in Barcelona") +
        xlab("Month") +
        ylab("Temperature (ºC)")

```

## Correlation between temperatures
```{r}
## Select year 1780 and 2016
meteo.extreme <- meteo.long[meteo.long$year==1780 |
                              meteo.long$year==2016,]

## Add two columns with temperature values
meteo.extreme.wide <- reshape2::dcast(meteo.extreme,
                                      month + season ~year, 
                                      value.var="temperature")
colnames(meteo.extreme.wide)[3:4] <- paste0("temperature.", colnames(meteo.extreme.wide)[3:4])

## Indclude a column with the difference in temperature
meteo.extreme.wide$temperature.difference <- meteo.extreme.wide$temperature.2016 - meteo.extreme.wide$temperature.1780

cor <- ggplot(meteo.extreme.wide, aes(temperature.1780, temperature.2016)) +
        geom_point(aes(color=season), size=3, alpha=0.7) +
        geom_text(aes(label=month))
cor

## ggrepel!! and add size as aesthetic!
cor <- ggplot(meteo.extreme.wide, aes(temperature.1780, temperature.2016)) +
          geom_point(aes(color=season, size=temperature.difference)) +
          ggrepel::geom_text_repel(aes(label=month))
cor

## Modify scales to be the same and change colors
cor <- cor + 
        scale_color_manual(values=scale_season, 
                           name="Season") +
        scale_size_continuous(name="Temperature \n Difference") +
        xlim(5, 25) + 
        ylim(5, 25)
cor

## Add line
cor <- cor + 
        geom_smooth(colour = "dark red", size = 0.5, alpha=0.3)
cor

## Add line for perfect correlation
cor <- cor + 
        geom_abline(intercept = 0, slope=1,
                    colour = "red", lty=2)
cor
```

If you want the points to be on top of the lines, just change the order of the geoms!

## Distribution of temperatures
```{r histogram}
ggplot(meteo.long, aes(temperature)) +
  geom_histogram() +
  facet_wrap(~season)
```

```{r density plot}
ggplot(meteo.long, aes(temperature)) +
  geom_density()

## But we want to see different distributions according to seasons!
ggplot(meteo.long, aes(temperature)) +
  geom_density(aes(color=season, fill=season), lwd=1, alpha=0.2) +
  scale_color_manual(values=scale_season) +
  scale_fill_manual(values=scale_season)

```

# Your turn! ChickWeight dataset
## Data
Load the dataset using `data("ChickWeight")` and explore the data with `View(ChickWeight)`. The description of the column values can be found typing `?ChickWeight`

## Exercise: Distributions
### Plot a histogram of weight distributions, faceting according to the Diet.
```{r}
ggplot(ChickWeight, aes(weight)) + 
  geom_histogram() +
  facet_wrap(~Diet, scales="free_y")

```

### Use geom_density to plot the density of weights coloring by Diet, and faceting by Time.
```{r}
ggplot(ChickWeight, aes(weight)) +
  geom_density(aes(color=Diet)) +
  facet_wrap(~Time, scales="free")
```




## Exercise: Boxplots
### Plot a boxplot of ChickWeight were x is Time and y is Weight.
```{r}
ggplot(ChickWeight, aes(factor(Time), weight)) + # We use factor() because x variable for boxplots cannot be numeric/continuous!
       geom_boxplot()
```

### Using the last boxplot, color groups according to the Diet.
```{r}
ggplot(ChickWeight, aes(factor(Time), weight)) + 
       geom_boxplot(aes(color=Diet))
```

## Exercise: Correlations
### Compare with a scatterplot the weigth of the chickens the first day (Time=0) and the last day (Time=21). Color the points according to the Diet

```{r}
chick <- ChickWeight[ChickWeight$Time==0 |
                       ChickWeight$Time==21, ]
chick.wide <- reshape2::dcast(chick, 
                              Chick + Diet ~ Time,
                              value.var = "weight")

ggplot(chick.wide, aes(`0`, `21`)) + # If names of the columns are numbers you can use `` to call them as objects.
  geom_point(aes(color=Diet))
```


# Your turn! If you have things you want me to explain or to try you can ask!