That's a pretty high correlation there skadistic. Now, if you'll excuse me if I get onto a high horse, but here goes.Correlation is not causation.
The admissibility of statistical evidence is governed by what is called the daubert principle. It's a court case you can look up. Its decision, in essence, is that statistical evidence must be developed in accordance to commonly accepted notions in the so-called expert community.So called. So a statistician,who can make statistics say what ever they want, can persuad the courts with manipulated numbers. What does that have to do with the fact that crack laws target crack and not the black race?
In anti-trust cases, this means that our models are quite elegant, and quite modern. This is mainly because the data available for such cases tends to be very well maintained and sufficient observations.Are crack laws anti-trust cases
For civil rights cases, ranging from housing price discrimination to incarceration rates to jim crow laws, the data happens to not be as robust, mainly because we can't increase the number of precincts in a jurisdiction, nor the number of voters. Bummer right? So they tend to use less sophisticated means. One of my responsibilities while attached to the CRD is to try and bring up the level of sophistication. One way I do this is by publishing articles with relevant caselaw.So how does housing discrimination add up to drug laws that target drugs and not races?
Basically, we have now derived that there is an 81% chance of a crack user being black, correct? I could say that the correlation is .81. I haven't inferred causality here. If we include socioeconomic status of "poor", I think you'd agree that the correlation would likely be similarly high.So. Does that show that crack laws target crack or blacks?
Now, if we run a simple linear regression, which is the court adopted method currently used, we're going to have a data problem. It is quite likely that the correlation between black and poor in the jurisdiction where the data is being pulled from is also quite high. This is called a problem of multi-collinearity.And that shows that crack laws target black instead of crack how?
Because there's no way to get around this problem currently, most civil rights litigation will accept correlation analysis, or alternatively, will except a simple bivariate regression equation with race as the explanatory factor. So long as the race variable is significant, in the eyes of the law, it does not matter if race is strongly correlated with education, political affiliation, income.Race is a factor? So the laws target blacks because there blacks or crackheads who happen to more likely be black?
In cases where I've prepared expert testimony for trial, or been said expert, this is the standard practice I've encountered. While I've tried to introduce more mathematical elegant models that show the interplay between multiple variables...I've either only been able to show that I cannot determine which variable is most causal (that they all need each other in order to be causal) or that the folks hearing the case just don't get my fancy statistics.Great your an expert. No show me the law about crack where it says blacks are the target and not crack.
Rather than tell folks that they're idiots and don't have a clue, I tried to explain how a legal court case involving racism claims is handled, especially the statistical evidence. I'm afraid here that the law (that which the supreme court justices ruled on) was quite much harsher on blacks than other racial groups, and thusly was overturned 7-2. The dissenting judges did not disagree with the notion that blacks had been more harshly sentenced than non blacks through this law. Rather, they dissented on the basis of a lack of procedure.
