57
One Proposed Solution
Apply weights due to be used late in study to the
most longterm experience
•In
the example, we would apply weights
–1/28,
2/28, 25/28
•Tends
to (appropriately) inflate variability of statistic at interim
analyses
•Intuitively
reasonable in that the results for the longest observations should
be more indicative of the future
–Similar
to imputing future observations