|
Activity Number:
|
190
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #300679 |
|
Title:
|
Simulation-Based Visualization of Inference Functions
|
|
Author(s):
|
Daeyoung Kim*+ and Bruce G. Lindsay
|
|
Companies:
|
The Pennsylvania State University and The Pennsylvania State University
|
|
Address:
|
Department of Statistics, 326 Thomas Building, University Park, PA, 16802,
|
|
Keywords:
|
Inference function ; Likelihood ratio statistic ; Score statistic ; Wald statistic ; Visualization ; Confidence sets
|
|
Abstract:
|
This paper presents a new simulation based methodology to visualize aspects of inference functions of statistical interest. Our simulation based methodology provides values for the parameters that picture the inference function surface around the estimator optimizing the inference function. Once done, one can picture the profile inference function confidence sets for multiple parameters of interest without further complicated optimization. Although this methodology is related to Fisher's concept of fiducial inference, it is here treated as merely a visualization device. Our method uses the same observed data without repeated sampling, which is also the case in Bayesian inference with Markov chain Monte Carlo algorithm. Moreover, our method does not require the specification of a prior distribution for parameters and a burn-in period for MCMC convergence.
|
