Picking your scientific nemesis: a practical guide
Not everyone needs a nemesis but I sure like having one!
Every now and then someone loudly and confidently asserts a scientific position with which you strongly disagree. Your disagreement is not just a knee-jerk, but because you’ve studied or thought about this particular issue on a level that warrants your own quiet, heroic confidence.
When this happens, it can really wind me up – in a constructive way! It’s actually motivated several papers in the past (either by me or by collaborators). Having a clear caricature of a position I think is wrong makes me worry that others might be persuaded, so I work really hard to my own understanding and arguments (and sometimes discover I’m wrong). For example:
My paper on the marginality principle was the result of a friend relaying some blanket advice from someone who confidently asserts wrong stuff all the time. Reading it made me think, ‘I disagree with this person about everything else but have always assumed this point is right. Given that this person is saying it, I have to work out what is wrong!’ I got talking with Tra and Maarten (plus a couple of others) and we wrote the paper.
Look at the quotes from an anonymous reviewer quoted in the introduction to this paper. When my colleague showed me these comments, we basically assumed it was correct that one-stage really was superior. So we got to work on comparing one-stage with two-stage. We realised that one-stage was more efficient only if it was fitting a different model that made extra assumptions. So differences aren’t due to one- or two-stage estimation procedures, but due to using different models (the comparison of number of stages is confounded). The person that I guess was the reviewer went on to publish a paper containing a really poor simulation study showing what they had already concluded. How? They found pretty much no difference between one- and two-stage meta-analysis procedures until they:
contrived a situation where a one-stage procedure makes a strong modelling assumption that (i) gains efficiency if right, (ii) leads to appreciable bias if wrong; and
studied case (i).
Art. We subsequently wrote this paper and my motivation was to avoid the other person’s paper persuading anyone (please don’t assume my co-authors’ motivation was the same as mine).
Other examples might make people’s identities obvious so I’m not going to list any more.
Criteria for a good nemesis
So clearly I find it motivating to picture someone as misleading unwitting readers, through ignorance or malevolence, and my task is to make sure the readers are witting (I don’t even need them to take my side, just understand the disagreement).
Without further ado, I present my non-exhaustive list of what to look for in a long-term nemesis.
1. Confidently asserts views regarding a topic on which you are expert
Of course, you do have to disagree with them (even if it’s on minor details). They should be so confident that they lack the ability to be circumspect about when they may be wrong. Or be unwilling to change their mind.
2. Their view/s should be (reasonably) specific to them
A lot of people loudly prefer R to Stata and will say why they prefer it as if it’s a fact or moral position rather than a preference (I think we know in our hearts that it’s really a matter of what we’re accustomed to). People will bemoan that the S- languages (SAS, Stata, SPSS, S-Plus, etc.) get used – they have a condition known as Rtockholm Ryndrome. You might find it annoying but it’s such a broadly held opinion that it’s hard to pick out one person as your nemesis from the crowd.
3. You have to be able to do something constructive with it
I’ve already said a couple of times why you might want a nemesis. But you need to ask yourself, is this relationship actually going anywhere? If you can’t use your disagreement constructively, you’re wasting your energy on this person. Actually the arguments about Stata vs. SAS vs. R are a bit like this. Like you’re just the old people walking to the supermarket and bickering: friends.
4. Your views should be (reasonably) specific to you
If almost everyone already agrees with you, there’s no point getting too worked up. Don’t choose the village idiot as your nemesis – to an outsider it’s hard to tell who is who, or you engaging may lend them credibility. I worried this would be a side-effect of the post-hoc empowerment episode (no you don’t get a link).
One (not village idiot!) example that got various people worked up was when Jeff Wooldridge, an influential econometrician, once expressed irritation that, if everyone used his definition of standard error (which he thought they should), then some of the ways people talk about standard errors would be wrong. Wooldridge defined the standard error as ‘an estimate of the SD of an estimator’*, and lots of people pointed out that SE has been been defined as ‘the SD of an estimator’ by the rest of the world for at least a century. Under his definition, making distinctions between things like the true, empirical and estimated SE makes no sense, but of course under the usual definition it does. He seems to have plenty of disciples though so maybe some even agree with him. Anyway, if lots of people are picking the argument, don’t try and claim this person as your nemesis.
*I wondered if he perhaps meant ‘an estimateor of the SD of an point estimator’. I’d still disagree but, as an estimator of the SD, it has properties that you can evaluate like bias vs. the actual SD of the estimator. An ‘estimate’ is just a number that’s spat-out. Yes yes, I know people say things like ‘biased estimates’ all the time and I probably have in the past.
5. A whole field can’t be your nemesis
We often see things that another field does and think it’s mad. You probably don’t want to turn that whole field into your nemesis. This will backfire and you’ll become that field’s Aunt Sally. It’s also useful to remember the Chesterton’s fence principle. I remember being stunned and outraged when I discovered certain fields using linear regression to estimate causal effects on binary outcomes. A few months ago I accidentally talked myself into it. Similarly, I think other fields are pretty aghast at biostatisticians’ love of using mixed models (I increasingly agree with their disapproval). Then I’ve also seen people criticise things in biostats where they clearly don’t understand the principles – effectively removing the fence without understanding its purpose
6. They should be (or seem to be) more senior, influential or powerful than you
It’s important to feel like the scrappy underdog. Those are are always the ones you root for in books and films. From outside, you don’t want to appear to be the powerful person punching down. Top tip: a good reason to dodge promotions and prizes is that it reduces the pool of possible nemeses you can choose from, which is obvs why I’ve never won anything. Eyes on the (real) prize!
7. Wrong about more than one thing!
It’s fine for someone to be a casual nemesis, but if you’re looking for something more long-term, you need to consider whether there are other things you can disagree on. Your relationship can evolve, rather than tailing off.
8. Think twice if you sit near them
Someone posted some slides on twitter from a small-time nemesis. I gently dunked on the slides, which seemed silly, and at the end said something snarky about the person’s previous record and some rat emailed them my tweet. I got a snotty email the next day from the person saying why their slides were right and that I was unprofessional. The awkward thing is they are in the same office space as me, and it’s a bit awkward when we’re both in the kitchen. We just avoid eye contact.
PS – the rat once sent me an email saying ‘I was about to tag one of your papers in a post but it looks like you’ve blocked me. I’d like to know why. Also, can you unblock me?’ Please have some dignity.
Advice: whatever you do, don’t tell them they’re that special person
Feelin’ shy? Don’t tell them your true feelings! Nemesis conversations with colleagues and collaborators are partly fun because they seem like a work conversation but actually have the same attraction as Real Housewives drama. The thing is, if you don’t want the drama of an actual feud with someone, probably don’t let them know you’ve chosen them. It’s supposed to be silly drama that stays in your head.
Fun fact
Diversion: I’ve had about 10 conversations with people about nemeses over the last 6–7 years. In at least six (maybe more), someone has named this one particular researcher as their nemesis, and each time for a different reason! I wondered if it would be constructive to let them know… then they did something to wind me up (not nemesis level), enough to see that they must know but care not.
What did I miss?
I’m sure I missed some important things. Comment!