Quote:
|
Originally Posted by chickentribs
If he didn't go into the study looking for the race correlation and it did catch him by surprise, it can be virtually impossible to go back on the same data and exclude factors with any reliablility - too many assumptions have to be made.
|
I'm not sure I understand here -- it's just a matter of including additional control variables in the analysis, perhaps one by one. Sample size limits the total number you can include at one time; a rule of thumb is you want to have at least 10 independent datapoints per variable included. That's undoubtedly the reason he didn't include everything on the first pass. It's a huge dataset.
Quote:
|
Thinking it through, it would actually be understandable that a black rookie would be "tested" more severely by black youth in these neighborhoods because they were seen many times as traitors to the race. It would also make sense that the black officer, especially a younger black man, could create substantial tensions within a city block.
|
Actually the study was much bigger than LA alone; it included a total of 189 U.S. cities in its restricted demographic dataset; the total dataset was many times this size (all U.S. districts with 100 officers or more). The effects were fairly uniform across cities; he did a few breakdowns in his tables for larger cities.
Quote:
|
If you haven't seen it yet, you should pick up Steven Levitt's book, Freakonomics. It is very layman so he doesn't substantiate proofs the way he would in texts, but it is a quick read and the questions he asks are interesting.
|
Thanks -- I'll definitely check it out, it sounds good, I still have a couple empty slots in my summer reading list. . . .