To the Editor: We welcome the opportunity to address briefly the interesting issues raised by Dr. Mammen and colleagues in response to our article. We disagree with the suggestion by Dr. Mammen et al. that our conclusions are the result of inappropriate statistical tests. We are aware that in statistical analyses involving small groups it is important when employing a two-sample t test that the data are normally distributed. For larger groups, however, it is not so critical, because of the central-limit theorem. With our data, for example, the group sizes were large enough and the distributions were sufficiently nonpathological that the two-sample t test worked well. We confirmed this by using the "double-bootstrap method" to carry out an extremely accurate version of the t test +(1). This calculation demonstrated that the ordinary two-sample t test works sufficiently well with these data to be considered an appropriate statistical test. Moreover, using a nonparametric test, as Dr. Mammen et al. suggest, yielded the same result, as use of the Wilcoxon rank sum test resulted in p=0.01 and p=0.005 for hopelessness and suicidal ideation, respectively.

We agree that the results of clinical association studies are frequently open to more than one interpretation and require replication, as we suggested. Our cross-sectional finding, that a greater score on the Reasons for Living Inventory may protect against the emergence of suicidal behavior during depression, is being tested in a prospective longitudinal study in which multivariate statistical analyses will be used. We look forward to completing and reporting on our findings in due course. In the meantime, we encourage researchers to consider developing clinical treatments to possibly enhance reasons for living during depression that may then be tested to see if they protect against the emergence of suicidal behavior.