Two Language Authentication
Last year I participated in a novel and exciting meta-analysis project called Many Analysts, One Dataset. A single dataset was given to researchers in over a dozen independent groups to analyze, with the hope of seeing just how divergent the analyses would be.
It was a great project, and they’re in the middle of a follow up now. The reason I mention it now is tangential. I initially tried to do the full analysis in Python, because Python is a lovely language which does not silently do unexpected things. Alas, the scientific python suite did not have a particular test statistic I needed and so I had to try again with R. When I re-implemented in R, though, I got a different result for one of the intermediate steps. After much fine-toothed-combing, I realized I’d made a simple error in the Python script that my basic unit tests hadn’t caught.
This experience led to a vow: for any important analysis I do, I will implement it in at least two languages.
Earlier today I came up with a name for this process: two language authentication.