Hi, I'm trying to find ressources on robust (approximated) Bayesian statistics, but I'm not finding what I'm looking for; maybe you can give me a hint where to look. Basically I'm looking for a way to get a BIC (Bayesian information criterion; Schwartz, 1978) for a model fit of robust methods. E.g. if I apply a robust regression (e.g., lmrob), is there a way to get (something like) a BIC for the model? For some regression models in R one can apply something like: stepAIC (mymodel.glm, k=log(n)) Or one can calculate the BIC based on the SSEs (sum of squared errors). I somehow fear/feel that the SSE-approach cannot be directly applied to robust methods as they use different measures to obtain their optimized estimates (e.g. least trimmed squares regression estimator). Can you give me hint where to look or how to think about this issue? Thanks! Sorry if I'm asking a painfully obvious or wrong question. Best regards, Stefan Herzog ------------------------------------------------------------- Dr. Stefan Herzog, Research Scientist Center for Cognitive and Decision Sciences Department of Psychology University of Basel Missionsstrasse 64A CH-4055 Basel Switzerland Tel +41 61 267 06 15 Fax +41 61 267 04 41 stefan.herzog at unibas.ch http://www.psycho.unibas.ch/herzog/
[RsR] Robust (approximated) Bayesian statistics: BIC of robust methods?
2 messages · Stefan Herzog, Elvezio Ronchetti
Hi, Check Machado(1993) Econometric Theory. All the best. Elvezio Ronchetti Dept. of Econometrics University of Geneva Blv. Pont d'Arve 40 CH-1211 Geneva SWITZERLAND e-mail Elvezio.Ronchetti at unige.ch tel +41 22 379 8131 tel (secr) +41 22 379 8229 Fax +41 22 379 8299 http://www.unige.ch/ses/metri/ronchetti/
Stefan Herzog wrote:
Hi, I'm trying to find ressources on robust (approximated) Bayesian statistics, but I'm not finding what I'm looking for; maybe you can give me a hint where to look. Basically I'm looking for a way to get a BIC (Bayesian information criterion; Schwartz, 1978) for a model fit of robust methods. E.g. if I apply a robust regression (e.g., lmrob), is there a way to get (something like) a BIC for the model? For some regression models in R one can apply something like: stepAIC (mymodel.glm, k=log(n)) Or one can calculate the BIC based on the SSEs (sum of squared errors). I somehow fear/feel that the SSE-approach cannot be directly applied to robust methods as they use different measures to obtain their optimized estimates (e.g. least trimmed squares regression estimator). Can you give me hint where to look or how to think about this issue? Thanks! Sorry if I'm asking a painfully obvious or wrong question. Best regards, Stefan Herzog ------------------------------------------------------------- Dr. Stefan Herzog, Research Scientist Center for Cognitive and Decision Sciences Department of Psychology University of Basel Missionsstrasse 64A CH-4055 Basel Switzerland Tel +41 61 267 06 15 Fax +41 61 267 04 41 stefan.herzog at unibas.ch http://www.psycho.unibas.ch/herzog/
_______________________________________________ R-SIG-Robust at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-robust