An Ode to Low R2

It’s the time of the year when many of us do their share of grading. In my case, it’s quantitative projects, and every time I’m impressed how much the students learn. One thing that annoys me sometimes is to see how many of them (MA students) insist on interpreting R2 in absolute terms (rather than to compare similar models, for instance). That’s something they seem to learn in their BA course:

[in this simple model with three predictor variables], we only explain 3% of the variance; it’s a ‘bad’ model.

I paraphrased, of course. But I started to like low R2: They are a testament to the complexity of humans and their social world. They are a testament to the fact that we are not machines, we are in the world where quantitative analysis is about tendencies. Just imagine a world in which knowing your age and gender I could perfectly predict your political preferences… So there you have it: low R2 are great!

Better the Average after Breaking the Chain

I have written before about habit formation and the longest chain or streak. The idea is simple: decide to do something (concrete) every day and measure how many days in a row you did that. Just don’t break the chain… go for the longest streak possible. I found this a useful approach for getting into daily habits.

There’s one problem, though: Once I have a long chain and break it, it seems hard to get started and going again. Say my daily task is to learn one new word. Every day I note for how many days I have been doing this. Say I’ve done 61 days in a row, and then break the chain. After a day or two (or so), I get back to studying new words. Now, after three days the incentives aren’t very strong. It’s still very far to get to the longest chain; if I don’t study, I only ‘lose’ three days — not much. (The incentives are quite difference when my chain is longest: now there’s everything to ‘lose’. This is why the whole thing works.)

What’s the solution? Don’t seek to do your best every time: go better than your average. This means that in addition to recording your current chain or streak, you also keep track of all the other chains. We still try not to break the chain, but there’s a secondary goal to better the average. (And if we use the geometric mean for the average, there’s even not that much information to be tracked)

Habit formation: What’s the longest chain?

It’s paradoxical… we humans are creatures of habits, yet it can be hard to form new habits. Here’s one thing I found useful: (1) deciding to do something regularly (e.g. daily), (2) tracking how many times in a row I did this. The motivational trick is trying to create the longest chain; if I fail to do whatever I’m tracking, I start at zero.

Now, the first step is important. It’s about a concrete, measurable goal. For example, do more reading doesn’t work; read one paper every Wednesday is. I found that less is more here: being too ambitious is more likely to lead to breaking the chain or just fatigue and motivational lows. By contrast, doing a little every day, or once a week, does add up… What is more, I schedule different things according to my week: weekends and weekdays are different, and days I spend commuting are different from days working from home.

For the second step, I also have clear rules (which vary according to the goal). For some, it’s every day, no excuses. For many there’s a list of acceptable excuses. For example, if I have to work later than a certain time, I allow myself to pick one of the remaining chains only. Or, for some goals I allow myself to skip once a week. I find these acceptable excuses useful to maintain the motivation of tracking chains, yet being able to adapt to real life. That said, I only permit one acceptable excuse in a row (irrespective of goal tracked).