Thursday, July 19, 2007

Race, Class, and Crime III: One possibility not considered in the previous two posts is that the correlation between race and crime is eliminated when family structure is controlled. I recoded the variable on family arrangements at age 16 to 1 for living with both parents and zero for everything else. Here are the multivariate results:

Logistic regression coefficients (t-statistic in parentheses)

Black .43 (3.29)
Mother's educ .00 (.15)
IQ -.02 (-.84)
Male 1.62 (15.25)
Age -.03 (-8.64)
Living with mom and dad at 16, -.48 (-4.28)

Number of valid cases 4,782

Family structure is related to being arrested, but it reduces the race-crime link only a little bit, according to GSS data.


  1. Random Guy8:00 PM

    Read and appreciated.

  2. Anonymous6:55 PM

    If it is a logistic regression, I think you mean z-statistic.

  3. Anonymous6:59 PM

    It would also be interesting to look at the interactions of race and family status.

  4. Anon: I use SPSS and it generates t-statistics.


Meta-analysis of clinical trials: Eat walnuts

I am always looking for easy eating choices that are good for you. This new meta-analysis of 26 clinical trials looked to see if walnuts ma...