|
Abstract:
|
In double blinded, randomized and placebo-controlled vaccine trials, Participants in vaccine and placebo groups will be block-randomized in groups of 3 at a ratio of 2:1 to receive vaccine or placebo, respectively. In the interim analysis, it may be desirable to look for preliminary evidence of treatment differences but, on either practical or ethical grounds, to avoid breaking the code to discover the sequence of treatment allocation employed. Due to blindness, each of group three data can be treated as a mixture model with two observations from treatment and one observation from control, however, the identities are unknown. In this talk we use an exponential tilting density model to link the treatment and control densities and study the maximum semiparametric likelihood estimation by theoretical derivations and numerical simulations.
|
