Comments on: Thought Crime Arrests 1928-1944

By: Sayaka

Sayaka — Mon, 29 Oct 2007 20:28:36 +0000

Although… you have to be able to compile the numbers by themselves. That is the most difficult task for historians, I guess..

By: Sayaka

Sayaka — Mon, 29 Oct 2007 20:25:30 +0000

I am not supportive of too much dependence of statistical data, either, but it is not difficult to avoid basic problems like the ones you discuss (e.g. the sample distribution), actually. When social scientists compile numbers by themselves today, they usually indicate the standard deviation, the number and cases of outliers, etc, to show an overall characteristic of the dataset. Numbers are very useful if you carefully use them.

By: K. M. Lawson

K. M. Lawson — Tue, 23 Oct 2007 23:43:50 +0000

That seems like a very sound policy you have.
The positive role they can play is huge, of course, when used carefully and claims are appropriately tempered.

By: Owen

Owen — Tue, 23 Oct 2007 17:19:46 +0000

I know this sort of problem well from my own work. I actually used quite a lot of charts and statistics that I compiled myself in my thesis and would often look back on them sometime after I’d done them and think, “that looks nice, but it doesn’t actually reveal anything at all”. In other cases the data that I had painstakingly and lovingly compiled just told me what I already knew and what historians had been saying in more general terms for a long time. Disappointing but perhaps useful in that it adds another layer to the empirical evidence for a certain picture of events. As you say, interpretation is the most tricky thing about this sort of historical data. My opinion is that you always need to provide some other sort of information alongside such statistical information, preferably information from narrative sources that can put the statistics in context, or which the statistics can provide quantitative support for.