arXiv:2404.17464v1 Announce Kind: cross
Summary: In most cancers analysis, general survival and development free survival are sometimes analyzed with the Cox mannequin. To estimate precisely the parameters within the mannequin, enough knowledge and, extra importantly, enough occasions must be noticed. In apply, that is typically an issue. Merging knowledge units from completely different medical facilities could assist, however this isn’t all the time doable attributable to strict privateness laws and logistic difficulties. Just lately, the Bayesian Federated Inference (BFI) technique for generalized linear fashions was proposed. With this technique the statistical analyses are carried out within the native facilities the place the info have been collected (or saved) and solely the inference outcomes are mixed to a single estimated mannequin; merging knowledge shouldn’t be vital. The BFI methodology goals to compute from the separate inference ends in the native facilities what would have been obtained if the evaluation had been based mostly on the merged knowledge units. On this paper we generalize the BFI methodology as initially developed for generalized linear fashions to survival fashions. Simulation research and actual knowledge analyses present wonderful efficiency; i.e., the outcomes obtained with the BFI methodology are similar to the outcomes obtained by analyzing the merged knowledge. An R package deal for doing the analyses is out there.
Supply hyperlink