I was recently reading about doing a side project during your PhD in a guest post on Pat Thompson’s blog patter  and at first I thought “only the really smart ones have time for that sort of thing” and quickly dismissed the idea.
I am not particularly clever when it comes to my field. I have always struggled with my degree  and though I, for some reason, feel certain that I can do a PhD, I also feel as if I have devote my full attention to it or risk failure. Clearly as I am writing this post I am at least entertaining the idea of a side project, so what changed my mind?
As I sit here each day struggling to implement Bayesian equivalents of t-tests and ANOVAs in R, I think: “how can I possibly contribute anything original in this area when it’s this difficult to work with the basic literature?”. Frankly, it’s slowly sapping my will to live. I contemplated writing a “Frequentist Vs Bayesian ANOVA” post but it just bored the hell out of me. Don’t get me wrong, I enjoy reading the papers and I believe my work will be valuable. I just can’t get excited about the methodological minutia of it all: I’m a big picture thinker and the nitty-gritty is just depressing.
Today, I was combing through papers  and, as my work is in Bayesian methods as a replacement for Frequentist methods, I can’t go two minutes without another debate about the reproducibility crisis in psychology; how p-values are the devil’s work; or how statistics is just a hot mess and nothing tells you anything any more. These debates have been happening for decades. To me it’s this that’s truly fascinating that there are papers from twenty years ago telling me how what they’re saying (in this case: p-values are the devil) isn’t new because it was in papers forty years prior to their paper.
The history of Bayesian statistics is pretty mysterious in my opinion. Firstly, if we all knew confidence interval’s and p-values were rubbish when they were invented why do we all use them and why are they still taught? Secondly, I have a personal goal to find out when precisely the standard form of Bayes’ Theorem appeared and who was responsible. People claim it was Laplace’s formulation but honestly I’ve read that paper and it did not, from my reading, explicitly state the standard formula where your prior could be whatever you wanted, Laplace used a uniform prior instead. I am also interested to see when proportionality came into it. When most people stop the story with Laplace, there’s still a big gap between his work and the modern .
So I am implementing a reward system: if I accomplish my weekly PhD goals I am allowed to pursue my own side interests as part of my work. This will hopefully keep me on my R grindstone and allow me to make real progress while giving me the freedom to pursue my own whimsical reasearch interests. Let’s see if this plan works, shall we?
- This is an excellent blog with a focus on academic writing which is particularly useful for doctoral and early career researchers.
- I do mean struggled. When I left with a 2:1 it was a very accurate picture of my potential.
- I guess you could call it procrastinating but I’m pretty sure I need to read them anyway so I’m going to call it “doing other work that’s not what you should be doing right now, but it’ll save you time later so that’s okay, right?”