I have a chip on my shoulder so why don’t we deal with a few things right off the bat:
- Statistical creativity is not about lying
- I am a statistician, not a mathematician
- Mentioning my job is a near-guaranteed conversation killer, so I’ll understand if you want to leave right now.
Are you still here? Cool! I commend you for your bravery and your taste in reading.
I have only seen one other profession that, when it comes up, stops people in their tracks as quickly as my own: economist. I think I understand why. On the one hand, statistics involves a lot of numbers. That triggers math phobia. On the other hand, statistical answers are usually some version of probably, thus scaring off math fans who like having a single right answer. But, in the end statistics are just a device for telling a story, a story that can be told with great panache or grinding boredom. I prefer panache.
So, if you imagine research as like writing a book, you’d not be too far off. Sure, such imagination won’t get you published, but you might be able to stop worrying about the best analogy for research. For all you writers or creators, here’s how creativity works in research and, more specifically, in statistics, my profession.
When funding runs low, researchers concoct an idea for what new knowledge would best further science or, more realistically, their career. This is a simple plot idea. The better researchers take this idea to a statistician (a biased viewpoint, I know) who acts as an editor, poking holes for the researcher to fill. At this phase, a statistician’s creativity consists of inventing what-if scenarios:
“What are you going to compare to your treatment? Are all patients the same? How much of my salary can you afford? Can you spell ‘chi-square’?”
After consulting a statistician and all sorts of wise and silly colleagues, this motivated research begins to construct a detailed research protocol or proposal. This is the book outline and, at this stage, the statistician becomes a co-author, contributing several key elements describing how the data will be analyzed, a justification for the chosen sample size, and various other parts requiring Greek letters and/or confusing non-medical terminology. Here, a statistician is not the primary author, but is an author with enough pull to be able to
sabotage save a brilliant project. Instead of poking holes, the statistician now becomes a problem solver and/or sales person.
“How will we make this work with so few subjects? You want to do an interim analysis and you just told me about it? Do you understand why I’m proposing such an ornate analysis? Again, how much of my salary can you afford?”
‘Writing’ the book
Assuming the project is funded (they rarely are), the next stage, running the study, is very much like writing the book, except the statistician has become a neurotic literary agent.
Patients Chapters come and go, each one contributing a plot twist to the final story. Meanwhile, the statistician looks in once in a while wondering when it will be finished and how much damage will have to be repaired later on. There’s very little creativity here, beyond whatever gossip is shared between statisticians (not that we’d do that).
“Are you going to actually get 50 subjects this year? That’s more missing data than I was expecting, can you fix it? Are you going to be done before I go on vacation?”
The main study is done and all the data is collected. Once, there was a plan and, if it hasn’t gone to hell, it certainly flirted with it a few times. At this point the statistician takes complete control, comparing plan to reality and trying to make them work well together. Although editing is not normally as creative a process, this is the single most creative part of a statistician’s job. Right now, after all the mistakes have been made, a statistician is asked to make lemon out of lemonade and, as you can tell, that requires some creative application of computers, Greek letters, and magic.
“The original analysis won’t work, you gave me 12 subjects when I needed 60, does this still answer your question? Shall I explain this statistic one more time? Since the numbers didn’t make sense to you, here’s this graph I spent last week creating so look at the pretty pictures, OK? What do you expect me to do with that since it doesn’t even look like data?”
Finally, it’s all done. Everything has been written except for one part: the blurbs. In the case of research, the blurbs are the final research article. It is a very important part, but 99% of the work goes into every other part of the study. For the statistician, there is one last burst of creativity because, for each article, journals expect a concise explanation of the statistical methods which usually means reducing 4 weeks of work to 2-3 sentences simple enough so that every reader can pretend to understand. Unfortunately, this tiny part is packed full of technical terms that almost every reader has heard, but very few truly understand. In short, this is a statistician’s chance to get revenge at the researcher’s never ending lectures on the biology of the whosie-whatsit that caused the worst cases of girandularium differensis commonly found in people with a specific genetic variant.
“Look, you don’t understand it and no one else will either and, although it says virtually nothing, this contains just enough confusing terminology so reviewers will be afraid to reject it. Trust me.”
We forgot. We always forget. The final draft is not the last part. For a book, you get one last chance to correct simple errors that crept in. For a research paper, you get to respond to reviewers
idiotic wise comments and tell them how ridiculous insightful they are and change everything to a different font, using Farsi letter instead of Greek so that the article says the same thing in completely different words.
“Yes, I promised it would work, but this time I really mean it. Would I use ‘heteroscedastic’ if I wasn’t truly serious?”
Since the first time through was so successful, why not do another one only this time on … [yawn] … huh? OK.
“So, how much of my salary can you cover this time?”