Fun with statistics

It is possible to "prove" practically anything by social science research:

Contrary to common beliefs, rising immigration levels do not drive up crime rates, particularly in poor communities, and Mexican-Americans are the least likely to commit crimes, according to a new study.
Robert Sampson, a sociologist at Harvard University who studied crime and immigration in 180 neighborhoods in Chicago over seven years, found that first-generation immigrants were 45 percent less likely to commit violent acts than third generation Americans.

Without access to Sampson's data, or without conducting any independent research of my own, I can spot what might be a few flaws in Sampson's study:

• First of all, why Chicago? Is the nation's third-largest city in some way representative of recent trends in immigration and crime? If it is not representative, then why was it chosen for the study?
• Did Sampson screen to separate legal immigrants from illegals in his data? Since the contemporary debate revolves around the problem of illegals, a failure to distinguish the two categories renders his study useless in discussing this crucial issue.
• Why the comparison between first-generation immigrants and third-generation Americans? While you're at it, why not compare them to DAR members?

Those are just a few quick thoughts about Sampson's findings. I'm sure Heather Mac Donald will look closely at the details of Sampson's study and figure out why it doesn't comport with her own work on the subject. Of course, when Mac Donald releases her own research, don't expect her study to be uncritically ballyhooed by the wire services. Only liberals get that treatment.