When comparing samples using multiple comparisons inflation of P values might be prevented using Bonferroni. The tradition was attackked by Perneger (BMJ 1998;316:1236-1238). From a Bayesian viewpoint what is really going on ?
When, in a paper (Kidney Biopsy as a predictor for renal outcome in ANCA-associated necrotizing glomerulonephritis. Kidney Int 1999;56:1751-1758) , multiple histological criteria are used to find which predicts best (in the paper: correlates "best" with...) the creatinine concentration one year later, I cannot see that Bonferroni should be used at all ? This is not really a hypothesis testing situation but explorative/descriptive - ?
Troels Ring, M.D
Department of Nephrology
Aalborg, Denmark
[log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|