JASP or not to JASP? Bayesian statistical methods for free.

There are two ways of doing statistical analysis. One that I call the Excel approach and the other that I call the Experts approach.

You use the former if you are a mouse user, and like to point and click at things without understanding what is happening. You can also forego control too.

And you use the latter for finer control over outputs. They require some understanding of the theoretical aspects of the analysis, and a lot more punching of the keyboard.

Introducing JASP

I’ve been experimenting a statistics software called JASP that aims to be a small statistics software — ‘an alternative to SPSS’. It falls in the realm of Excel users but has the power of Experts underneath. What? Is THAT POSSIBLE?

JASP has only a few, well… FOUR, categories of analysis — t-tests, anova, regression, and frequencies — and this is fine for most of the use cases. I find it particularly interesting for EDA (Exploratory Data Analysis) where you just want to look at the data your working with.

Everything is visual and done with the mouse, from loading data, to clicking through the options of the analysis you want to perform.

The output is beautiful but limited to HTML with the images embedded as png data:uris. It would be great to export the results for use in academic papers — eps or pdfs would be great for LaTeX. This makes JASP a little limiting.

What about Mondrian?

In any case, JASP is still in its infancy and will evolve rapidly. As for exploratory data analysis software there are alternatives. One of my favourites is Mondrian. Mondrian is old as it started in 1997, latest stable is from 2011 and latest beta is from 2013. Mondrian doesn’t have the beautiful graphics of JASP either. Sometimes simple things are the best and Mondrian just works.

But isn’t JASP just a gui for R?

The cool thing about JASP is that all the clicks and clanks are just a way to pass parameters to R functions. The backend is R and therefore you can do everything the hard way if you really need to improve on the existing functions. For example, sometimes you need to make your plot scales logarithmic. There is no way to do that in JASP but if you use the R function underlying it… you solve your problem — and you can export PDFs for paper production.

The R scripts that power JASP are available at Github. If you want to use them in your own advanced work in R just go ahead and read them.