Tuesday, October 18, 2016

The Posts Not Finished

The ones which sob and moan in the middle of the night when I can't sleep, the ones which I tossed into the deep snow where they wander shoeless and coatless, in cold pain, looking for mama.  The ones which had so much care and work and effort spent on their nursing but which, nevertheless, I cruelly rejected and abandoned.

Those posts.

How does that beginning sound?  I veered off the topic there, because the posts I want to talk about here, the ones never finished, are not of the emotional sob-story kind.  No, they are statistical posts, based on an enormous amount of work by me:  Calculations and spreadsheets and all sorts of other boring yet electrifying crap.

Why these posts have never stepped into the limelight of this humble blog vary.  For example, I worked long on a post about the US Congress, about how representative it is for various demographic groups (such as, say, comparing the percentage of Latinas in the Congress to their percentage  in the US population).

But I gave up on it because of all sorts of tricky statistical problems, such as trying to find out if Latino Congress members are also counted again in the race categories, and if so, what I should do about it.  I got extremely uninformative answers to my queries from those who had compile some statistics I tried to use.

And then I wondered if anyone would be even interested in the findings (which suggest, as one would expect, that white Anglo men are over-represented, but which also suggest that not all minority groups are under-represented to the same extent, or at all, and that women, in general, are under-represented within all racial or ethnic categories).

Then there are the police shootings data, the fatalities among black men, white men and other population groups.  I spent quite a bit of time analyzing the Washington Post surveys for 2015 and 2016, going through their data case by case, calculating all sorts of averages and percentages.

And I may still write up that work.  But when to post it?  The time never appears to be right, because the work I have done is not emotional work.  It doesn't seem fit to post it when yet another black man is killed by the police, because it would sound like an instrument in the orchestra playing a different tune from all the others.  To post it at any other time would limit its exposure.

Then there are all the questions I have about those data sets.  How are they verified?  Why is the race of so many who died not recorded?  Is it because the data comes from newspaper articles?  If so, how many cases are not reported at all or reported wrong?  Whose reports are used when deciding if the killed person* was armed or not?

Finally, the data sets themselves seem to show a lot of short-term variation.  The relative number of Hispanic men killed by the police in 2015 was considerably higher than the relative numbers in the first half of 2016, though the relative numbers of black men killed remained fairly stable**.  It would be good to understand that, and other data characteristics better before writing about the surveys.

So are you sufficiently bored yet?  How about this topic for a post:  Suppose that before you are born you are told that one third of your life will be spent on practicing being dead.  Wouldn't you feel cheated out of all those years?  But we don't think of sleep that way.


* Those killed persons were, by the way, overwhelmingly men, especially in the unarmed category.

** And how do those who create the data set decide if the killed man was white or black or Latino?  Latinos can be either white or black, too, or can belong to other racial categories.