Posted: Dec 29, 2010 11:14 am
Thommo wrote:Just to dive in with the devil's advocate stuff: The sample size (27 murders, with treated data) is far too small to state with any degree of confidence that no significant difference exists between murder rates for biological children and non biological children, only that no significant difference was found for it. If (and this is the mighty big supposition you were discussing) there was a general pattern of abuse towards non-biological children as compared to towards biological children I don't think it would be reasonable to say that this finding was significantly against the trend - you'd likely need far bigger sample to make it improbable (depending on effect size).
Unfortunately the sample size is small given that it's a fairly rare event that they're looking at but I suppose there are a few things we need to keep in mind:
1) the numbers are big enough to reach significance in the chi square test
2) the data were taken using information spanning from 1965-2009, which is a huge population to take from (and to sort through), so it will be costly and very time consuming trying to do research on larger populations
3) the research supporting the Cinderella effect uses comparable (and sometimes smaller) sample sizes, and
4) the research supporting the Cinderella effect doesn't take into account the confound of increased violence in stepfamilies (which makes their conclusions very shaky).
So I do agree that before we can make any substantive claims about the "demise of a paradigm" we need to replicate this research, check that it generalises to other populations, etc, but on the other hand, the research supporting the Cinderella effect wasn't exactly concrete so it wouldn't take much to topple it.