[生物书合集].Advances.in.Clinical.TBiostatisti

[生物书合集].Advances.in.Clinical.TBiostatisti

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
For example, the dose for each patient might be chosen to minimize the posterior expected loss with respect to the loss function L ( x , N )= d { u , p ( x , N )} or L ( x , N )= m ( x , g ) for some choice of metrics d and m defined on the unit square and S ± G , respectively. Thus, patients might be treated at the mean, median, or mode of the marginal posterior distribution of the MTD, corresponding to the respective choices of loss function L ( x , N )= ( x ² g ) 2 , L ( x , N )= j x ² g j , and L ( x , N )= I (0, q ) ( j x ² g j ), for some arbitrarily small positive constant q . Instead of minimizing the posterior expected loss, dose levels can be chosen so as to minimize the loss function after substituting an estimate for N . Consequently, given the data from k patients, one might estimate N as ˆ N k and administer to the next patient the dose x k þ 1 ¼ arg min x e S f L ð x ; ˆ N k Þg : In the remainder of this section we describe various loss functions that have been discussed in the literature concerning cancer phase I clinical trials. Since the primary statistical aim of a phase I clinical trial is to
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.
Ask a homework question - tutors are online