|
Abstract:
|
Computation of the full data maximum likelihood survival parameter estimates under the Cox proportional hazards assumption proves computationally cumbersome, since the number of parameters is approximately equal to the number of observations. We show that there exists a low-dimensional functional sub-space which contains the complete data maximum likelihood survival function. By optimizing within this fixed-dimensional sub-space, the computational burden of optimizing the likelihood is significantly reduced. In this presentation, we compare the full data maximum likelihood estimates to the usual partial likelihood. We will also compare the running time between our method and the typical full data maximum likelihood computation.
|
