ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
ST 520 Statistical Principles of Clinical Trials - NCSU Statistics ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
CHAPTER 7 <strong>ST</strong> <strong>520</strong>, A. TSIATIS and D. Zhang<br />
|Tn34| = |2<br />
� �1/2 640 × 639<br />
(.633 − .715)| = 2.94 ∗ .<br />
640 + 639<br />
Clearly, treatments 2 and 4 are the better treatments, certainly better than control. The only<br />
controversial comparison is treatment 2 versus treatment 3, where, we may not be able to conclude<br />
that treatment 2 is significantly better than treatment 3 because <strong>of</strong> the conservativeness <strong>of</strong> the<br />
Bonferroni correction<br />
7.4 K-sample tests for continuous response<br />
For a clinical trial where we randomize patients to one <strong>of</strong> K > 2 treatments and the primary<br />
outcome is a continuous measurement, then our primary interest may be to test for differences in<br />
the mean response among the K treatments. Data from such a clinical trial may be summarized<br />
as realizations <strong>of</strong> the iid random vectors<br />
(Yi, Ai), i = 1, . . ., n,<br />
where Yi denotes the response (continuously distributed) for the i-th individual and Ai denotes<br />
the treatment (1, 2, . . ., K) that the i-th individual was assigned. Let us denote the treatment-<br />
specific mean and variance <strong>of</strong> response by<br />
and<br />
Note:<br />
E(Yi|Ai = j) = µj, j = 1, . . .,K<br />
var(Yi|Ai = j) = σ 2 Y j , j = 1, . . .,K.<br />
1. Often, we make the assumption that the treatment-specific variances are equal; i.e. σ 2 Y 1 =<br />
. . . = σ2 Y K = σ2 Y , but this assumption is not necessary for the subsequent development.<br />
2. Moreover, it is also <strong>of</strong>ten assumed that the treatment-specific distribution <strong>of</strong> response is<br />
normally distributed with equal variances; i.e.<br />
(Yi|Ai = j) ∼ N(µj, σ 2 Y ), j = 1, . . .,K<br />
Again, this assumption is not necessary for the subsequent development.<br />
PAGE 109