The generalizability of the results is limited by the nature of our
community sample, which comprised urban young adults. Additionally,
despite the relatively large sample, the typically low
conditional probability of PTSD resulted in a small number of
PTSD cases (n 121), limiting statistical power and precluding
more complex analytic models that include all suspected codeterminants.
The test of a gender difference in the sensitizing effect of
prior assault on the PTSD response of a nonassault trauma (a test
that does not address statistically the distinctiveness of the effect
for subsequent nonassaultive vs. assaultive violence) reached significance
at the conventional level of .05. However, the
gender difference in the specificity of the sensitization by type of
prior event, as well as by type of subsequent event was significant
only at a less conservative alpha, which has been suggested for
interactions in logistic regression (Hosmer & Lemeshow, 1989).
Replications in larger samples are needed. With respect to codeterminants,
race and family socioeconomic status indicators were
unrelated to the probability of PTSD in this sample (Breslau,
Wilcox, et al., 2004). Major depression or anxiety disorders can
not be considered potential confounders in the gender–PTSD differential
(given that gender is fixed at birth), although history of
these disorders increases the risk for PTSD. The general caveat
about the reliance on retrospective data in epidemiological studies
also applies to these results. Data on exposure and PTSD are based
on respondents’ accounts of their lifetime experiences and thus the
accuracy of the data might be less than desired. However, the
young age of the cohort would tend to attenuate this concern.