In his 1925 book “Statistical Methods for Research Workers” Ronald Fisher wrote:
The value for which P =.05, or 1 in 20, is 1.96 or nearly 2 ; it is convenient to take this point as a limit in judging whether a deviation is to be considered significant or not. Deviations exceeding twice the standard deviation are thus formally regarded as significant.
Since then, a P-value less than 0.05 (about two standard deviations, or sigmas) has been nearly universally used as the test for rejecting the null hypothesis of studies, trials, and experiments in the life sciences. The cutoff is arbitrary, although any P-value cutoff would be, since statistical analysis of an experimental readout fundamentally reports only the likelihood that the result was due to random chance. It’s up to the scientist to assume a level of chance deemed acceptable, and it’s more palatable if everyone agrees to the same playing field rather than applying their own risk tolerance.
The inspiration for the name of this blog isn’t from the arbitrary nature of this cutoff (a necessary evil) but rather the nature in which it became the standard. There was no consortium of leading figures gathered to determine the most appropriate treatment; rather, in the space of two sentences, Fisher moved the characterization of P=0.05 from “convenient” to “formally regarded as significant”. I don’t mean to disparage Fisher here. He was already a highly influential polymath with major contributions in the field of statistical genetics. In fact, there was probably no better person to make this decision in 1925, and a decision did need to be made. For me, the takeaway is that asserting preferences as rules to make progress is sometimes necessary. In science, it probably happens more than most people realize.
The purpose of writing this blog is multi-faceted. Some posts will explore interesting datasets, or try to answer a nagging question, or advance my own knowledge of methods in data science by developing models in my professional areas of expertise. Others will be more of a discussion without hard evidence and data. However, the principal goal is to make a space to discuss, analyze, and explore topics that I think are significant and interesting. Hopefully you do as well.