Welcome

Welcome to this Data Visualisation workshop at the Schiermonnikoog Life History Theory Course. This workshop helps you in making graphs in R via ggplot2 and hopefully drives the point home that, above all else, show the data. ggplot2 is a visualization package for R, written by Hadley Wickham (Wickham 2016), and it is simply fantastic. At lightning speed, we’ll first cover R and RStudio, before we go on to learning how to visualize our data (beautifully) in R. You can switch to different topics using the navigation bar on the left.

Just to give you a glimpse of the amazingness of ggplot, see what a few lines of code in R can produce:

ggplot(mpg, aes(x=cty, y=hwy, colour=drv)) + geom_point() + geom_smooth() + theme_minimal()

Above all else, show the data

In his book The Visual Display of Quantitative Information (1983), visual artist Edward Tufte “Above all else, show the data”. This quote has always resonated with me. Showing people all the data, allows insights but also scrutiny. Most graphs problematically hide most of the data (e.g., by showing only means). It is my strong belief that the replication crisis would not have been as severe, if our cultural norms of doing science included the practice of visualisation in which all or at leat most of the data are shown.

Why are visualizations important?

Graphs can change the world, as Florence Nightingale showed:

The importance of visualizations is further evident from Anscombe’s quartet:

And from the datasaurus dozen:

And from:

Why R?

Some reasons to use R:

It’s free

You can easily generate stuff:

# Flipping a coin ten times
sample(c("Head", "Tails"), size = 10, replace = TRUE, prob = c(0.5, 0.5))

##  [1] "Tails" "Head"  "Head"  "Head"  "Tails" "Tails" "Head"  "Tails" "Tails"
## [10] "Head"

You can do ‘standard’ analysis, like linear regression:

model <- lm(hwy ~ cty + drv, data = mpg)
summary(model)

## 
## Call:
## lm(formula = hwy ~ cty + drv, data = mpg)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -4.7830 -0.9041 -0.3083  0.8973  5.0223 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  1.99883    0.46622   4.287 2.66e-05 ***
## cty          1.19859    0.03087  38.826  < 2e-16 ***
## drvf         2.22365    0.27027   8.227 1.42e-14 ***
## drvr         2.12501    0.33314   6.379 9.70e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 1.494 on 230 degrees of freedom
## Multiple R-squared:  0.9379, Adjusted R-squared:  0.9371 
## F-statistic:  1157 on 3 and 230 DF,  p-value: < 2.2e-16

You can also do fancy “state-of-the-art” analysis stuff, for example:

network analyses
mixed models
missing data imputation
bayesian analyses
making statistical webapplications

You’ll work more reproducibly:

You’ll learn about the wonderful world of coding:

passed_the_test <- function(grade) {
    if(grade > 5.5){
        print("You passed the test")
    }
    else {
        print("You failed the test")
    }
}
passed_the_test(8)

## [1] "You passed the test"

You can easily and beautifully visualise stuff:

ggplot(mpg, aes(x=cty, y=hwy, colour=drv)) + geom_point() + geom_smooth() + theme_minimal()

You can not-always-easily-but-beautifully visualise stuff:

Source: https://spatial.ly/2012/02/great-maps-ggplot2/

You’ll learn the virtue of patience as R frustrates you:

References

Wickham, Hadley. 2016. Ggplot2: Elegant Graphics for Data Analysis. 2nd ed. Cham, Switzerland: Springer International Publishing. http://www.springer.com/br/book/9780387981413.

‘Above all else, show the data’ - workshop Schiermonnikoog