STATISTICAL ANALYSIS OF CHANGES IN THE DECISIVENESS OF
We used a logistic model to examine whether the decisiveness of general acceptance has changed over time. The outcome variables are whether the element of evidence was found unreliable and whether the challenged evidence was excluded. The key explanatory variables are (1) whether the evidence was rated unfavorably on general acceptance and (2) whether the evidence was rated unfavorably on any other reliability factor. The two variables were interacted with time to test whether the decisiveness of general acceptance and that of the other reliability factors have changed. A positive coefficient on the interaction between time and a reliability factor means that the probability that the element of evidence will be found unreliable when it is rated negatively on that factor (holding its rating on the other factors constant) increased during the time period. As before, dummy variables for case type, substantive area of evidence, and appellate circuit were included to control for a variety of evidence-specific characteristics that remain constant over time. The model was also estimated allowing for correlation among the errors for elements of evidence coming from the same opinion.
The logit model has the same form as the model described in Appendix B, but here the explanatory variables are
and the other variables are as defined to Appendix B.
Table D.1 reports the results. The second and third columns show the results for whether the evidence elements are found unreliable. As can be seen, after Daubert, general acceptance is a good predictor of whether evidence was found reliable when ratings on other reliability factors are held constant. (The coefficient 3.5699 has the expected sign and is statistically significant at 10 percent.) Prior to Daubert, the coefficient on general acceptance (the sum of 3.5699 and