Re: Re: Re: Re: Re: Chernobyl effect on Swedish educational attainment.
Anyone notice the lovely chartsmanship?
Even with the chartsmanship in place the variation was a little difficult to discriminate. Take the chartsmanship away and ...
This looks like an excellent exercise for a budding epidemiologist.
"Here, run the numbers on this."
They get to spend lots of time calculating lots fo numbers and in the end they get to pat themselves on the back.
Whenever someone compiles educational information down into a point (i.e. 3% decrease), my alarms go off. People can be points, but make them points not amalgamations. Take those 1176 kids that fell into the target group and plot their grades. Now do the same with the "control" group. Overlay them and get your "Gut" feel.
I may have missed it, but did they compare against 1980 and 1990 for the same regions?
Re: Re: Re: Re: Re: Re: Chernobyl effect on Swedish educational attainment.
I think there was mention somewhere that they could only go to 1988 birth year since, at the time they gathered the data, that was the last birth year for which the later results were available. Bit it is a working paper and I think they have plans to milk any forthcoming interest.
The low number in the affected group compared to the trojan number was notable. Worse the seemingly unpublished figures (and very dodgy concept though at first sight it seems neat?) for comparing same sea siblings 2 years either side of 1986 birth date ranges. That seems to offer an opportunity for some very low numbers with all the documented control constraints applied.
Also I seem to recall seeing a comment about the student assessment method changing that year.
I don't know what to think about it really. Somewhere between the chartmanship and the dataset there is a truth to be told. I'm just not sure where it is.