Abstract:
|
The traditional power prior proposed by Ibrahim and Chen (2000) does not account for different (heterogeneous) data types in the historical and current data, and it is generally not usable in such settings. Thus, a new type of power prior needs to be developed in situations where the historical and current data are of different data types, which we call the scale transformed power prior (straPP). In this talk, we develop the straPP for situations where the current (historical) data are right censored time-to-event data and the historical (current) data have a different data type, such as continuous, binary, count, etc.. The scale transformed power prior is constructed so that the information matrix based the current data likelihood is appropriately scaled by the information matrix from the power prior, thereby shifting the scale of the parameter vector from the historical data to the new data. Several examples are presented to motivate the scale transformation and several simulations and real data sets involving time-to-event studies are presented to show the advantages in performance of the scale transformed power prior over the power prior and other priors.
|