Introducing cannonball - Tools for teaching statistics
I’ve put my first R package on GitHub!
cannonball and contains a couple of functions that I use for teaching;
perhaps others will follow.
Make sure you have the
Then load it and install
To use it, load the package as per usual:
Overview of the functions
plot_r(): Draw scatterplots with the same correlation coefficient
People seem to like this function from my blog post What data patterns can lie behind a correlation coefficient?. Specify the number of observations and a desired sample Pearson correlation coefficient, and out come 16 rather different looking scatterplots conforming to these criteria:
For more details, type in
?plot_r at the R prompt.
clustered_data(): Simulate data from a cluster-randomised experiment
Cluster-randomised experiments are experiments in which whole groups
of participants (e.g., entire classes) are necessarily assigned to the same
condition. If the data from such experiments are analysed as though the
participants were assigned to the conditions individually (e.g., by
running a t-test on the individual data points), the false positive rate
can go through the roof.
This function simulates data from such an experiment and allows
you to specify unequal cluster sizes (via the
I mostly use this function in a simulation to illustrate the effects of clustering on p-values. With a null effect, you’d expect only 5% of the p-values to be lower than 0.05. Let’s see what happens when you analyse the individual data from a cluster-randomised experiment using a t-test:
The false positive rate is now through the roof (33%).
Graphically checking model assumptions
See the full-fledged tutorial for these functions.
Glad you asked! Julian ‘Cannonball’ Adderley is one of my favourite alto saxophone players (check out his solos on Autumn Leaves (from around 2’03”; Somethin’ Else) or Freddie Freeloader (6’22”; Kind of Blue)!) and he was a consummate jazz educator to boot.