As you all know, I’m a big fan of reproducible research. Most people have an idealized conception of scientific research and that, by and large, it’s all about the pursuit of truth. Sadly that’s not always the case. The problem isn’t that all problematic results are faked; sometimes human nature is enough to corrupt the results.
Over at the Pacific Standard, Jerry Adler has an interesting article that is nominally about research in Psychology but has implications for other scientific research, including, astonishingly, Mathematics.
It’s astounding how often data is massaged to support a preconceived hypothesis. The article describes a study that used real data to “prove” an preposterous result. The study was up front about the silliness of the result and was making a point about how easily data can be manipulated.
Not all of these problems would be solved by reproducible research but having the data and the specific manipulations that it was subjected to available would go a long way towards ending this problem. Read the article to get an idea of the extent of the problem. After reading it, see if you don’t agree that journals should insist that submitted articles provide access to the type of data envisioned by reproducible research. Perhaps, then, the Reproducibility of important research in fields such as bio-medicine would do a bit better than one in ten.