Far East Journal of Theoretical Statistics
Volume 4, Issue 2, Pages 347 - 362
(December 2000)
|
|
OUTLIER
DETECTION WITH FRACTIONAL BAYES FACTOR
Choon-Il Park (Korea)
|
Abstract: In this paper, we modify the fractional Bayes factor (FBF;
O’Hagan [J. Royal Stat. Soc. Series B 57 (1995),
99-138]) with the generalized Savage-Dickey density
ratio (Verdinelli and Wasserman [J. Amer. Stat. Assoc.
90 (1995), 614-618]) to overcome the unknown constant
problem in Bayes factor from the improper priors. This
modified FBF are applied to detecting outlier in
random effect model with a mean-shift model. Finally,
we have a simulation with a hypothetical data set
including an outlier and analyze a real data set. |
Keywords and phrases: fraction Bayes factor, mean-shift
model, outlier detection, Savage-Dickey density ratio,
important sampling method. |
|
Number of Downloads: 5 | Number of Views: 809 |
|