Setting up
Back in 2008, hospitals were reporting an increase in injuries related to playing Nintendo Wii (http://www.telegraph.co.uk/news/uknews/1576244/Spate-of-injuries-blamed-on-Nintendo-Wii.html). These injuries were attributed mainly to muscle and tendon strains. A researcher was interested to see whether these injuries could be prevented. They hypothesized that a stretching warm-up before playing Wii would help lower injuries, especially for athletes because their regular activity makes them more flexible. She took 60 athletes and 60 non-athletes (factor 1: athlete), and within this group half did a 5-minute stretch routine before playing/watching whereas the other half did not (factor 2: stretch). The outcome was a pain score out of 10 (where 0 is no pain, and 10 is severe pain) after playing Nintendo Wii for 4 hours (outcome: injury). Let’s conduct a factorial ANOVA to test whether being an athlete and/or doing stretching before playing could prevent injury.
library(tidyverse) # what would we do without it?
wii <- ________
### Let's revise the commands in the past recitations. Let's create a summary table with the descriptive statistics of the dataset (mean, min, max, sd, se) for each group.
```r
wiiSummary <- _______ %>%
__________
Why don’t we plot this?
interactionPlot <- ggplot(_______, aes(x=_______, y=_______, # let's put the athlete factor on the x-axis
# Remember -- the next three arguments of the aes function will tell ggplot the aesthetics of the points and lines we are going to plot:
color=_______, # `color` wants the *variable name* which we want the contour of the geometries below to change by
fill=_______, # `fill` wants the *variable name* which we want the filling of the geometries below to change by
group=_______)) + # `group` wants the *variable name* which we want to group the dataset by
geom_????() + geom_?????() # plot overlaying point and line geometries
interactionPlot # call the plot to see it in the pane
What does the plot suggest? It is ordinal/monotonic or disordinal/non-monotonic?
Modelling
Let’s build an ANOVA model to see if injury is predicted by either/both factors. But first – what is the format of the dataset? It is ready to be part of the ANOVA calculations? Yes, it is, since it is in long format!
The formula will have now two predictors on the right of the tilde ~ operator. Let’s take a look at the formula help on the right to see how to properly construct the formula with two (or more) predictors.
model <- aov(_____ ~ _____ ? ___, data=____) # we should name it so to call it
model # like this
_____(model) # let's see what the model says and let's interpret the ANOVA table together.
What does the model tell us? What’s the the interpretation of the results?