Online Program Home
My Program

Abstract Details

Activity Number: 253 - Contributed Poster Presentations: Section on Statistical Computing
Type: Contributed
Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Computing
Abstract #329893
Title: Caveats on Data Cloning
Author(s): Brian Zaharatos*
Companies: University of Colorado Boulder
Keywords: data cloning; identifiability; maximum likelihood estimation; Bayesian
Abstract:

For the maximum likelihood estimator (MLE) to be unique, the parameter must be both identifiable and estimable. A parameter is identifiable if there is a one-to-one correspondence between parameter values and density functions. A parameter is estimable if the likelihood function has a unique mode. The method of data cloning has been proposed as a way to diagnose structural deficiencies-such as non-identifiability and inestimability-in a model. In this paper, we discuss cases in which this might be problematic, including cases where the number of clones required to detect model deficiencies may be impractically large. We then provide guidelines for more sound identifiability diagnostics.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program