AN evaluation of non-iterative methods for estimating the linear-by-linear parameter of ordinal log-linear models

Eric J. Beh, Thomas B Farver

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

Summary Parameter estimation for association and log-linear models is an important aspect of the analysis of cross-classified categorical data. Classically, iterative procedures, including Newton's method and iterative scaling, have typically been used to calculate the maximum likelihood estimates of these parameters. An important special case occurs when the categorical variables are ordinal and this has received a considerable amount of attention for more than 20 years. This is because models for such cases involve the estimation of a parameter that quantifies the linear-by-linear association and is directly linked with the natural logarithm of the common odds ratio. The past five years has seen the development of non-iterative procedures for estimating the linear-by-linear parameter for ordinal log-linear models. Such procedures have been shown to lead to numerically equivalent estimates when compared with iterative, maximum likelihood estimates. Such procedures also enable the researcher to avoid some of the computational difficulties that commonly arise with iterative algorithms. This paper investigates and evaluates the performance of three non-iterative procedures for estimating this parameter by considering 14 contingency tables that have appeared in the statistical and allied literature. The estimation of the standard error of the association parameter is also considered.

Original languageEnglish (US)
Pages (from-to)335-352
Number of pages18
JournalAustralian and New Zealand Journal of Statistics
Volume51
Issue number3
DOIs
StatePublished - Sep 2009

Fingerprint

Log-linear Models
Evaluation
Maximum Likelihood Estimate
Common ratio
Natural logarithm
Categorical variable
Nominal or categorical data
Odds Ratio
Contingency Table
Iterative Procedure
Standard error
Newton Methods
Iterative Algorithm
Parameter Estimation
Quantify
Log-linear models
Scaling
Calculate
Evaluate
Estimate

Keywords

  • Association model
  • Contingency table
  • Newton's method
  • Ordered categorical variable
  • Orthogonal polynomials

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

@article{ca8515fa86fa4602b612a6434277aa8c,
title = "AN evaluation of non-iterative methods for estimating the linear-by-linear parameter of ordinal log-linear models",
abstract = "Summary Parameter estimation for association and log-linear models is an important aspect of the analysis of cross-classified categorical data. Classically, iterative procedures, including Newton's method and iterative scaling, have typically been used to calculate the maximum likelihood estimates of these parameters. An important special case occurs when the categorical variables are ordinal and this has received a considerable amount of attention for more than 20 years. This is because models for such cases involve the estimation of a parameter that quantifies the linear-by-linear association and is directly linked with the natural logarithm of the common odds ratio. The past five years has seen the development of non-iterative procedures for estimating the linear-by-linear parameter for ordinal log-linear models. Such procedures have been shown to lead to numerically equivalent estimates when compared with iterative, maximum likelihood estimates. Such procedures also enable the researcher to avoid some of the computational difficulties that commonly arise with iterative algorithms. This paper investigates and evaluates the performance of three non-iterative procedures for estimating this parameter by considering 14 contingency tables that have appeared in the statistical and allied literature. The estimation of the standard error of the association parameter is also considered.",
keywords = "Association model, Contingency table, Newton's method, Ordered categorical variable, Orthogonal polynomials",
author = "Beh, {Eric J.} and Farver, {Thomas B}",
year = "2009",
month = "9",
doi = "10.1111/j.1467-842X.2009.00549.x",
language = "English (US)",
volume = "51",
pages = "335--352",
journal = "Australian and New Zealand Journal of Statistics",
issn = "1369-1473",
publisher = "Wiley-Blackwell",
number = "3",

}

TY - JOUR

T1 - AN evaluation of non-iterative methods for estimating the linear-by-linear parameter of ordinal log-linear models

AU - Beh, Eric J.

AU - Farver, Thomas B

PY - 2009/9

Y1 - 2009/9

N2 - Summary Parameter estimation for association and log-linear models is an important aspect of the analysis of cross-classified categorical data. Classically, iterative procedures, including Newton's method and iterative scaling, have typically been used to calculate the maximum likelihood estimates of these parameters. An important special case occurs when the categorical variables are ordinal and this has received a considerable amount of attention for more than 20 years. This is because models for such cases involve the estimation of a parameter that quantifies the linear-by-linear association and is directly linked with the natural logarithm of the common odds ratio. The past five years has seen the development of non-iterative procedures for estimating the linear-by-linear parameter for ordinal log-linear models. Such procedures have been shown to lead to numerically equivalent estimates when compared with iterative, maximum likelihood estimates. Such procedures also enable the researcher to avoid some of the computational difficulties that commonly arise with iterative algorithms. This paper investigates and evaluates the performance of three non-iterative procedures for estimating this parameter by considering 14 contingency tables that have appeared in the statistical and allied literature. The estimation of the standard error of the association parameter is also considered.

AB - Summary Parameter estimation for association and log-linear models is an important aspect of the analysis of cross-classified categorical data. Classically, iterative procedures, including Newton's method and iterative scaling, have typically been used to calculate the maximum likelihood estimates of these parameters. An important special case occurs when the categorical variables are ordinal and this has received a considerable amount of attention for more than 20 years. This is because models for such cases involve the estimation of a parameter that quantifies the linear-by-linear association and is directly linked with the natural logarithm of the common odds ratio. The past five years has seen the development of non-iterative procedures for estimating the linear-by-linear parameter for ordinal log-linear models. Such procedures have been shown to lead to numerically equivalent estimates when compared with iterative, maximum likelihood estimates. Such procedures also enable the researcher to avoid some of the computational difficulties that commonly arise with iterative algorithms. This paper investigates and evaluates the performance of three non-iterative procedures for estimating this parameter by considering 14 contingency tables that have appeared in the statistical and allied literature. The estimation of the standard error of the association parameter is also considered.

KW - Association model

KW - Contingency table

KW - Newton's method

KW - Ordered categorical variable

KW - Orthogonal polynomials

UR - http://www.scopus.com/inward/record.url?scp=69949096932&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=69949096932&partnerID=8YFLogxK

U2 - 10.1111/j.1467-842X.2009.00549.x

DO - 10.1111/j.1467-842X.2009.00549.x

M3 - Article

AN - SCOPUS:69949096932

VL - 51

SP - 335

EP - 352

JO - Australian and New Zealand Journal of Statistics

JF - Australian and New Zealand Journal of Statistics

SN - 1369-1473

IS - 3

ER -