Biomedical Engineering Reference
In-Depth Information
FIGURE 11.2: MSE for Month=1
Biases as depicted in Figure 11.1 can have a fundamental impact on the statistical
significance of tests of hypotheses on the associated parameters. Considering, for
example, the treatment coecient of 1 from the case of \Month=3" and p=0.1,
the downward biases from \Cox.Right" and \Cox.Mid" changed the conclusion on
treatment eect from signicance in \IntCox" to nonsignicance in both \Cox.Right"
and \Cox.Mid," as seen in Figure 11.3.
Figure 11.3, three histograms are plotted. The top reects \IntCox," the middle
reects \Cox.Right," and the bottom reects \Cox.Mid" from the 1,000 simulations.
The three solid lines on the histograms are the 2.5%, 50%, and 97.5% quantiles. It
can be seen that the 95% CI for \IntCox" is (0.055, 1.036), which corresponds to a
statistically significantly treatment effect because the CI does not cover zero. How-
ever, the 95% CI are (0.108,0.909) and (0.033, 0.927) for \Cox.Right" and \Cox.Mid,"
 
Search WWH ::




Custom Search