I’ve long been a fan of “How to shoot yourself in the foot” jokes. Having shot myself in the foot with different programming languages — particularly with C — I was thinking about how one might shoot oneself in the foot with various statistical approaches. So, here we go…
Subjective Bayesianism (de Finetti): Believe you’ve shot yourself in the foot. Or not. Don’t tell me what to believe!
Objective Bayesianism (Jeffreys): Build the uniquely-appropriate gun for shooting yourself in the foot. Oh, you want to shoot your left foot? That’s a different gun.
Fisherian p values (Fisher): Buy a gun to shoot yourself in the foot. You don’t know why it works, and the salesperson can’t tell you, but everyone uses this gun. Shoot yourself in the hand.
Frequentism (Neyman): Build a gun that, when you pull the trigger over and over, more often than not will shoot yourself in the foot. You can pull the trigger, but you can’t say whether you’ve shot yourself in the foot.
Likelihoodism (Edwards): Borrow a gun from a Bayesian.
And finally, because it all started with programming languages….
How to shoot yourself in the foot with R: Post on R-help. Someone there will be happy to shoot you.
If you have more, post them in the comments!
The APS Observer has just published a profile of JASP, a graphical user interface designed to make statistics easier. It includes Bayesian procedures by means of the R and the BayesFactor package. From the article:
JASP distinguishes itself from SPSS by being as simple, intuitive, and approachable as possible, and by making accessible some of the latest developments in Bayesian analyses. At time of writing, JASP version 0.6 implements the following analysis tools in both their classical and Bayesian manifestations:
- Descriptive statistics
- t tests
- Independent samples ANOVA
- Repeated measures ANOVA
- Linear regression
- Contingency tables
Read more at the APS observer.
Felix Schönbrodt and I have been working on an R package called BayesFactorExtras. This package is designed to work with the BayesFactor package, providing features beyond the core BayesFactor functionality. Currently in the package are:
- Sequential Bayes factor plots for visualization of how the Bayes factor changes as data come in: seqBFplot()
- Ability to embed R objects directly into HTML reports for reproducible, sharable science: createDownloadURI()
- Interactive BayesFactor objects in HTML reports; just print the object in a knitr document.
- Interactive MCMC objects in HTML reports; just print the object in a knitr document.
I anticipate releasing this to CRAN soon.