sampling bias

["Mark J. Chaffin" <Mark-Chaffin@xxxxxxxxx>]

Regarding the Rind, et al. study and the issue that this represents a less
biased sample because it exclusively used college students, it is important
to note that recent data presented by Duncan has found that 1) abused
students have markedly higher freshman year college drop-out rates than
non-abused students, 2) freshmen with more abuse drop out at higher rates
than those with less abuse, and 3) among abused freshmen, severity of PTSD
symptoms is significantly associated with dropout.  I think it is difficult
to avoid the fact that college students are a somewhat "purified" sample
which will disproportionately represent the most healthy members of the
abused population, and to recognize the limits of statistical control or
cross-study comparisons in dealing with this problem.  It is important to
bear in mind that college student samples are longitudinal, with high levels
of both pre-inclusion and post-inclusion attrition, and that cross-sectional
samples of this population are vulnerable to some inherent biases.  And this
is not to mention the interpretative bias, which was the real problem in the
Rind et al. study.

Mark Chaffin, Ph.D.
Center on Child Abuse and Neglect
University of Oklahoma Health Sciences Center