So I’ve always hated the idea of blogging. The pressure of coming up with something interesting to say once a month, a week, or more often, seems an added pressure that most of us could do without. But an academic blog could and should be exempt from such pressure. At its core it should be useful. It should serve a purpose. It should impart ‘wisdom’, whatever that may mean.
I’ve now started my own website, which of course you are aware of if you are reading this. Having my own website is something I’ve been meaning to do for years, indeed a few people have been trying to convince me to do this for a long time, to provide more information on my software programs and methods that I’ve developed. Don’t I feel important…Yes my posts will predominantly be filled with sarcasm.
I’ve written a lot of different software programs, all in Stata, that people seem to find useful. This includes anything from simulating and modelling complex survival models, to extended funnel plots in meta-analysis, to joint modelling longitudinal and survival data. My latest release is for multi-state modelling, which is a project I’m particularly proud of. But the big project I’m currently working on, megenreg, is my biggest challenge so far, and I think the most useful. I’ll write some future posts on it later.
My plan for this blog is to write some posts which hopefully researchers find useful. A lot of my posts I expect will be driven by the many interesting questions I get sent in emails, motivated by people using my software, or having attended one of my short courses. The more interesting, and more widely useful of these I will turn into fully fledged tutorials on my software pages.
I’ve now meandered my way to the crux of this post - developing user-written software in the modern age of biostatistics. Doesn’t that just sound impressive. My website will centre on the software I develop, and have developed over the past few years.
The best piece of advice I could give any young aspiring biostatistician is to develop and release software. It doesn’t really matter in what language. Yes R, SAS and Stata are the big hitters in biostats, but if you can write code in one language, you can do it in any. My point is that methods development in statistics often suffers from the fundamental problem that people think it’s too complicated, and not useful. I could not agree more. The fault lies entirely with us, the methodological statistician. If you publish a methodological paper in Biometrics, JRSS Series B, Stats in Med or any other technical journal, and your paper does not come with the crucial phrase of “we provide user-friendly software”, then quite frankly, your paper is pointless. Many will disagree with that, but then I will disagree with them.
Write software, document it, and publicise it. Anyone can work in biostats research, but if you want to do something really useful, then write and release a software package. I’ve built a career on it.