3 reasons to subcontract the statistical analyses in the scientific projects
January 20, 2018
Outsourcing statistics in demanding research projects will save money, help keep the deadlines & improve the integrity of the research.
As a founder of EcoStat (statistical consulting), my motivation to write this post is selfish and commercial. Yet also honest. I actually believe that outsourcing statistics is a healthy practice, especially in the modern day science culture with stiff deadlines and heavy proposal writings (otherwise I would not have started the company).
Budgeting the statistics already into the project proposal will save time, money, and also lead to better scientific practices. That statement is based on my own observations while working and collaborating with other researchers.
Saving time and money
What made the human race as prosperous and happy as it is today? The TRADE. Specialisation and collaboration is always superior to self-sufficiency.
There are couple of key skills needed in ecological science, one of them is statistics and numerical ecology. Not liking numbers is quite common among biologists, and to overcome this disaffection (only so much to be able to complete the statistical analyses by oneself) can take lot of time. And loads of unnecessary despair. Unnecessary, because statistics is like a language – something you get very good at when you do it all the time, but always painful and full of mistakes, when you do it only at utmost need.
Co-authorship as a trade
Typical solution is to ask a good friend or colleague to do it for you. This is already in spirit of trade.
But it also has downsides:
a) completing your papers depends on someone else having time for you;
b) doing it only for co-authorship can also mean that you won't be very dedicated to it.
Hired help saves time and money
For a large enough group, own statistician is probably best solution.
For smaller projects and groups, the need for statistics is only sporadic, once the data is ready.
In that case, outsourcing statistics will be better solution:
a) it saves money (you only pay for the actual work done),
b) help keeping deadlines, because there is a professional ready for you at the time of need, and this will also be a person fluent in statistics, working on similar tasks on a daily basis - hence faster in completing the tasks.
Improving integrity of research
There’s another interesting thing I’ve noticed with people and data.
When it is your project, data and hypothesis, you tend to test different statistical methods, until you get the result you need (and you are very likely to trust the results that agrees with your expectation). But that is a poor scientific practice.
Curiously, giving the data for someone neutral and professional, makes it psychologically much easier to accept the fact that the results did not support the hypothesis (and publish it as such); because you can skip the responsibility of doing the hard decision (“killing your darlings”) yourself.
People also tend to postpone reaching the conclusions, when it is their own data (because they don’t like the results, or because they’re not completely sure, if they’ve tried the best methods). A statistician will look at the data without any expectations, and make the decision based on the best practice and without any emotions. Which means, whether or not you liked the results, you can move on, and use the time to write and think about why the results were what they were.
And that’s what the science should really look like.