|
Activity Number:
|
465
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 7, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences
|
| Abstract - #300265 |
|
Title:
|
What is the Chance that a Death Sentence Will be Overturned?
|
|
Author(s):
|
Yves Chretien*+ and Andrew Gelman and Alex Kiss
|
|
Companies:
|
Columbia University and Columbia University and Sunnybrook and Women's College Hospital
|
|
Address:
|
1828 Lerner Hall, New York, NY, 10027-8311,
|
|
Keywords:
|
Bayesian ; hierarchical ; death penalty ; multistage ; survival ; missing data
|
|
Abstract:
|
When a death sentence is imposed in the U.S., a defendant is accorded the right of appeal at three successive stages. We use a multistage Bayesian hierarchical logistic regression to estimate the overall probability of a death sentence being reversed as a function of individual-level and state-level covariates. Estimation of the reversal rate is complicated by censoring and missingness of the data. Since most of the sentences imposed during our study period (1973-1995) were handed down in recent years, and since the cases may take more than 10 years to move through the judicial system, our data on the fate of each death sentence are heavily right-censored. Also, due to the unavailability of certain data from the second stage of the appeals process, we have an informative pattern of missingness. By using a multistage survival model in conjunction with our logistic model we can impute values for missing data points and therefore surmount these problems. This allows us to estimate the overall probability that a death sentence imposed during the study period would ultimately be reversed.
|
