|
Activity Number:
|
82
- Contributed Poster Presentations: Government Statistics Section
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Government Statistics Section
|
|
Abstract #309805
|
|
|
Title:
|
Sample Size Selection in Continuation-Ratio Logit Mixed Effects Models
|
|
Author(s):
|
Andrew M. Raim* and Thomas Mathew and Kimberly Sellers
|
|
Companies:
|
U.S. Census Bureau and University of Maryland, Baltimore County and Georgetown University
|
|
Keywords:
|
Generalized Linear Mixed Models;
Laplace Approximation;
Longford Approximation;
Quadrature;
Wald Test;
General Linear Hypothesis
|
|
Abstract:
|
The continuation-ratio logit (CRL) is a parameterization of the multinomial distribution which naturally models a sequence of L binary trials carried out until either the first success is observed or until failures occur in all L trials. The probability of success at each stage can be linked to a regression which may contain both fixed and random effects. This work explores computational methods to support power studies under CRL models with fixed effects or both fixed and random effects. Such power studies can be used to justify the selection of sample size when planning a future experiment. Our motivating application is an experiment designed for the 2020 decennial census, where CRL provides a model for nonresponse in a sequence of enumeration attempts to a household. The goal is to measure the effectiveness of a new training module for enumerators that provides guidance when contacting non-English speaking households.
|
Authors who are presenting talks have a * after their name.
