Econstudentlog

Preliminary data on the blood glucose/mental performance ‘study’

If you don’t know what I’m talking about, here’s the introduction.

I haven’t done as many sessions as I’d have liked, but at this point n is equal to 50 so I figured I might as well give you a scatter plot with the performance data so far:

Without the 2100+ performance at 17 mmol/l (the far right data point) R^2 would be 0,1463 – so n is still way too low to draw any conclusions. Perhaps aside from the fact that I don’t think the pattern looks completely random.

I’ve become aware of the fact that there are just loads of omitted variables here (nearby road work done with extensive use of pneumatic drills being one of the major ones in the beginning) and it would take a lot of data to take them all into account.

I’ve also realized by now that the tactics trainer performance is not a super great tool to pick up on variation in mental ability, though I maintain it’s not completely crazy to use it as a proxy. A significant number of the problems during a session are either repeats or quite similar to other problems solved in the past, and I remember those patterns just as well with a high blood glucose as with a lower one. So most of the variation in performance is around a set baseline, and how much I deviate from that baseline depends on how many ‘new’ problems – where I actually do have to think a bit – are introduced during a session. My performance is quite sensitive to the type of problems presented during a session and to which degree new problems/themes are introduced – the performance can easily vary with 200 points or more if I do two sessions ten minutes apart.

November 12, 2012 - Posted by | Chess, Diabetes, Personal

1. Yep, the limited number of non-trivial non-similar chess problems and the annoying human traits of memory and learning are likely to introduce a upward drift over time – very good for survival, bad for modeling human responses. Damn adaptive humans ðŸ™‚

One thing that jumps out at me from the scatter plot is the empty area b/w 4 and 6 mmol/L and the 1900 skill lvl. It seems that your skill in that blood sugar level is notably higher – based, of course, on 8 observations.

I’d be curious to see how the data points would be clustered by a k-means model. I suspect something like this or something like this, although you never know – the human brain seems to have much better heuristics for one-variable regression than for one-variable clustering, especially if you look for more than 2 clusters.

On a totally unrelated note, any comments on the scrapping of the fat tax in Denmark? It’s truly fascinating (to me) what a team of faceless bureaucrats can conjure up as “good policy”, and then backtrack on it because of bad outcomes that were totally predictable – higher prices, people complaining, people switching to cheaper alternatives (both at home and across the border in Germany), etc. What was the thinking there – people will not notice, people will keep quiet, people are too stupid to adapt? This is not a dig at Danish politicians/bureaucrats in particular, by the way – in fact, they deserve more credit than most others for at least learning from a mistake, and folding rather than doubling down to save face.

Comment by Plamus | November 13, 2012 | Reply

• The high average performance around 4-6 mmol/l may not be random even if it’s so far based on few observation – I often feel much, much sharper when my blood glucose is at that level than I ‘normally do’ (most of the time my blood glucose is not in that range) and I’ve always felt that way. I sometimes like to think my performance around that point is close to how I’d perform more generally in the hypothetical case where I wasn’t a diabetic, which is a depressing notion.

I don’t really follow the political debate much, but one thing to note regarding the scrapping of the fat tax is that the government revenue generated by the tax will now instead be generated by an increase in the lowest marginal income tax rate. Of course I didn’t like the fat tax but in terms of ‘fairness’ I’m not sure this change is a slam dunk improvement of affairs. Though it most certainly is a more politically palatable solution.

Comment by US | November 13, 2012 | Reply