Thomas William Yee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Thomas William Yee has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
- Chenchen Ma
Center for Forecasting Science, Academy of Mathematics and Systems Science, Chinese Academy of Sciences
2 shared publications - Alan H. Welsh
Research School of Finance, Actuarial Studies and Statistics, The Australian National University, Canberra, Australia
1 shared publication - Alec G. Stephenson
National University of Singapore
1 shared publication - Alfian Futuhul Hadi
Department of Mathematics University of Jember Jember Indonesia
1 shared publication Ana Clara F. Becker
Laboratório de Ecologia e Conservação Marinha, Instituto do Mar, Universidade Federal de São Paulo, Santos, São Paulo, Brazil
1 shared publication- Anthony Rodgers
The George Institute for Global Health Level 8, 55 Botany Street, Randwick NSW 2031
1 shared publication
Latest Publications
- 1 December 2025
- Published by Institute of Mathematical Statistics in The Annals of Applied Statistics
- Vol. 19 (4) , 3045-3070
Heaping and seeping, GAITD regression and doubly constrained reduced-rank vector generalized linear models in smoking studies
Large-scale health surveys suitable for addiction studies furnish self-reported data that consequently suffer from a form of measurement error called heaping, which statisticians have been grappling with for decades. Also known as digit preference, the aberration is often characterized by spikes at multiples of 10 or 5 upon rounding. To date, methods and software for heaped (and seeped) data have been largely wanting. Identifying three generic problems for simple addiction studies, we solve them by a newly developed technique, called Generally Altered, Inflated, Truncated and Deflated (GAITD) regression, for counts applied to a recent National Health and Nutrition Examination Survey data set. In conjunction, we propose the class of doubly constrained reduced-rank VGLMs whereby the reduced-rank regression is afforded structure by way of linear constraints—this allows further simplification of the dimension reduction. We determine the distribution of smoking initiation age (SIA) and its association with tobacco consumption and smoking duration; for example, is a lower SIA associated with higher tobacco consumption later in life? Is higher SIA associated with shorter smoking duration among quitters? Together, GAITD regression and DRR-VGLMs hold promise for heaped and seeped data.
- 1 December 2025
- Published by Institute of Mathematical Statistics in The Annals of Applied Statistics
- Vol. 19 (4) , 3045-3070
Heaping and seeping, GAITD regression and doubly constrained reduced-rank vector generalized linear models in smoking studies
Large-scale health surveys suitable for addiction studies furnish self-reported data that consequently suffer from a form of measurement error called heaping, which statisticians have been grappling with for decades. Also known as digit preference, the aberration is often characterized by spikes at multiples of 10 or 5 upon rounding. To date, methods and software for heaped (and seeped) data have been largely wanting. Identifying three generic problems for simple addiction studies, we solve them by a newly developed technique, called Generally Altered, Inflated, Truncated and Deflated (GAITD) regression, for counts applied to a recent National Health and Nutrition Examination Survey data set. In conjunction, we propose the class of doubly constrained reduced-rank VGLMs whereby the reduced-rank regression is afforded structure by way of linear constraints—this allows further simplification of the dimension reduction. We determine the distribution of smoking initiation age (SIA) and its association with tobacco consumption and smoking duration; for example, is a lower SIA associated with higher tobacco consumption later in life? Is higher SIA associated with shorter smoking duration among quitters? Together, GAITD regression and DRR-VGLMs hold promise for heaped and seeped data.
- 1 November 2024
- Published by Institute of Mathematical Statistics in Statistical Science
- Vol. 39 (4) , 568-588
Generally Altered, Inflated, Truncated and Deflated Regression
Models such as the zero-inflated and zero-altered Poisson and zero-truncated binomial are now well established. We review important ideas behind the three operators involved (as well as deflation and modification) and other incremental extensions. We propose a super mixture model that unifies alteration, inflation, truncation and deflation for counts, given a 1- or 2-parameter parent or base distribution. Since all of the operators except for truncation have both parametric and nonparametric variants, it is necessary to consider a total of seven different scenarios. Highlights of this paper include the following: (i) the mixture distribution is exceeding flexible, accommodating up to seven modes; (ii) under-, equi- and over-dispersion can be handled using a negative binomial (NB) parent, with underdispersion handled by a novel Generally-Truncated-Expansion method; (iii) under- and over-dispersion are studied holistically in terms of the the four operators previously referred to; (iv) while generally-altered regression explains why observations are there, generally-inflated regression accounts for why they are there in excess, and generally-deflated regression explains why observations are not there. The important application of heaped and seeped data from retrospective self-reported surveys is briefly mentioned. The GAITD-NB has potential to become a Swiss army knife for analyzing count responses.
- 1 November 2024
- Published by Institute of Mathematical Statistics in Statistical Science
- Vol. 39 (4) , 568-588
Generally Altered, Inflated, Truncated and Deflated Regression
Models such as the zero-inflated and zero-altered Poisson and zero-truncated binomial are now well established. We review important ideas behind the three operators involved (as well as deflation and modification) and other incremental extensions. We propose a super mixture model that unifies alteration, inflation, truncation and deflation for counts, given a 1- or 2-parameter parent or base distribution. Since all of the operators except for truncation have both parametric and nonparametric variants, it is necessary to consider a total of seven different scenarios. Highlights of this paper include the following: (i) the mixture distribution is exceeding flexible, accommodating up to seven modes; (ii) under-, equi- and over-dispersion can be handled using a negative binomial (NB) parent, with underdispersion handled by a novel Generally-Truncated-Expansion method; (iii) under- and over-dispersion are studied holistically in terms of the the four operators previously referred to; (iv) while generally-altered regression explains why observations are there, generally-inflated regression accounts for why they are there in excess, and generally-deflated regression explains why observations are not there. The important application of heaped and seeped data from retrospective self-reported surveys is briefly mentioned. The GAITD-NB has potential to become a Swiss army knife for analyzing count responses.
- 1 January 2024
- Published by Instats
Vector Generalized Linear and Additive Models
This workshop provides a comprehensive introduction to Vector Generalized Linear and Additive Models (VGLMs and VGAMs), equipping researchers with advanced skills for analyzing complex data structures often encountered in the social, health, and natural sciences. Participants will gain practical experience using the VGAM package in R for data analysis, enhancing their ability to apply sophisticated yet easy-to-use methods for their own research projects.
- 1 January 2024
- Published by Instats
Vector Generalized Linear and Additive Models
This workshop provides a comprehensive introduction to Vector Generalized Linear and Additive Models (VGLMs and VGAMs), equipping researchers with advanced skills for analyzing complex data structures often encountered in the social, health, and natural sciences. Participants will gain practical experience using the VGAM package in R for data analysis, enhancing their ability to apply sophisticated yet easy-to-use methods for their own research projects.
- Open Access
- 26 July 2023
- Published by Springer Nature in Japanese Journal of Statistics and Data Science
- Vol. 6 (2) , 827-846
Taylor’s power law and reduced-rank vector generalized linear models
Taylor’s power law (TPL) from empirical ecological theory has had many explanations proposed for its widespread observation in data. We show that the class of reduced-rank vector generalized linear models (RR-VGLMs) for coupling two parameters from a statistical distribution linearly together creates hybrid models that satisfy TPL or very similar. These include the RR-negative binomial, RR-inverse Gaussian and RR-generalized Poisson distributions. Some advantages of RR-VGLMs include the handling of covariates and an implementation exists in the form of theVGAMpackage. The software is demonstrated to show how these models may be fitted conveniently.
- 26 July 2023
- Published by Springer Nature in Japanese Journal of Statistics and Data Science
- Vol. 6 (2) , 827-846
Open AccessTaylor’s power law and reduced-rank vector generalized linear models
Taylor’s power law (TPL) from empirical ecological theory has had many explanations proposed for its widespread observation in data. We show that the class of reduced-rank vector generalized linear models (RR-VGLMs) for coupling two parameters from a statistical distribution linearly together creates hybrid models that satisfy TPL or very similar. These include the RR-negative binomial, RR-inverse Gaussian and RR-generalized Poisson distributions. Some advantages of RR-VGLMs include the handling of covariates and an implementation exists in the form of theVGAMpackage. The software is demonstrated to show how these models may be fitted conveniently.
- Open Access
On the Hauck-Donner Effect in Wald Tests: Detection, Tipping Points, and Parameter Space Characterization
The Wald test remains ubiquitous in statistical practice despite shortcomings such as its inaccuracy in small samples and lack of invariance under reparameterization. This paper develops on another but lesser-known shortcoming called the Hauck--Donner effect (HDE) whereby a Wald test statistic is not monotonely increasing as a function of increasing distance between the parameter estimate and the null value. Resulting in an upward biased $p$-value and loss of power, the aberration can lead to very damaging consequences such as in variable selection. The HDE afflicts many types of regression models and corresponds to estimates near the boundary of the parameter space. This article presents several new results, and its main contributions are to (i) propose a very general test for detecting the HDE, regardless of its underlying cause; (ii) fundamentally characterize the HDE by pairwise ratios of Wald and Rao score and likelihood ratio test statistics for 1-parameter distributions; (iii) show that the parameter space may be partitioned into an interior encased by 5 HDE severity measures (faint, weak, moderate, strong, extreme); (iv) prove that a necessary condition for the HDE in a 2 by 2 table is a log odds ratio of at least 2; (v) give some practical guidelines about HDE-free hypothesis testing. Overall, practical post-fit tests can now be conducted potentially to any model estimated by iteratively reweighted least squares, such as the generalized linear model (GLM) and Vector GLM (VGLM) classes, the latter which encompasses many popular regression models.
Open AccessOn the Hauck-Donner Effect in Wald Tests: Detection, Tipping Points, and Parameter Space Characterization
The Wald test remains ubiquitous in statistical practice despite shortcomings such as its inaccuracy in small samples and lack of invariance under reparameterization. This paper develops on another but lesser-known shortcoming called the Hauck--Donner effect (HDE) whereby a Wald test statistic is not monotonely increasing as a function of increasing distance between the parameter estimate and the null value. Resulting in an upward biased $p$-value and loss of power, the aberration can lead to very damaging consequences such as in variable selection. The HDE afflicts many types of regression models and corresponds to estimates near the boundary of the parameter space. This article presents several new results, and its main contributions are to (i) propose a very general test for detecting the HDE, regardless of its underlying cause; (ii) fundamentally characterize the HDE by pairwise ratios of Wald and Rao score and likelihood ratio test statistics for 1-parameter distributions; (iii) show that the parameter space may be partitioned into an interior encased by 5 HDE severity measures (faint, weak, moderate, strong, extreme); (iv) prove that a necessary condition for the HDE in a 2 by 2 table is a log odds ratio of at least 2; (v) give some practical guidelines about HDE-free hypothesis testing. Overall, practical post-fit tests can now be conducted potentially to any model estimated by iteratively reweighted least squares, such as the generalized linear model (GLM) and Vector GLM (VGLM) classes, the latter which encompasses many popular regression models.
- Open Access
- 23 February 2022
- Published by The Royal Society in Proceedings Of The Royal Society B-Biological Sciences
- Vol. 289 (1969) , 20211834
Ecosystem engineer morphological traits and taxon identity shape biodiversity across the euphotic–mesophotic transition
The euphotic–mesophotic transition is characterized by dramatic changes in environmental conditions, which can significantly alter the functioning of ecosystem engineers and the structure of their associated communities. However, the drivers of biodiversity change across the euphotic–mesophotic transition remain unclear. Here, we investigated the mechanisms affecting the biodiversity-supporting potential of free-living red coralline algae—globally important habitat creators—towards mesophotic depths. Across a 73 m depth gradient, we observed a general decline in macrofaunal biodiversity (fauna abundance, taxon richness and alpha diversity), but an increase in beta-diversity (i.e. variation between assemblages) at the deepest site (86 m depth, where light levels were less than 1% surface irradiance). We identified a gradient in abundance decline rather than distinct ecological shifts, driven by a complex interaction between declining light availability, declining size of the coralline algal host individuals and a changing host taxonomy. However, despite abundance declines, high between-assemblage variability at deeper depths allowed biodiversity-supporting potential to be maintained, highlighting their importance as coastal refugia.
- 23 February 2022
- Published by The Royal Society in Proceedings Of The Royal Society B-Biological Sciences
- Vol. 289 (1969) , 20211834
Open AccessEcosystem engineer morphological traits and taxon identity shape biodiversity across the euphotic–mesophotic transition
The euphotic–mesophotic transition is characterized by dramatic changes in environmental conditions, which can significantly alter the functioning of ecosystem engineers and the structure of their associated communities. However, the drivers of biodiversity change across the euphotic–mesophotic transition remain unclear. Here, we investigated the mechanisms affecting the biodiversity-supporting potential of free-living red coralline algae—globally important habitat creators—towards mesophotic depths. Across a 73 m depth gradient, we observed a general decline in macrofaunal biodiversity (fauna abundance, taxon richness and alpha diversity), but an increase in beta-diversity (i.e. variation between assemblages) at the deepest site (86 m depth, where light levels were less than 1% surface irradiance). We identified a gradient in abundance decline rather than distinct ecological shifts, driven by a complex interaction between declining light availability, declining size of the coralline algal host individuals and a changing host taxonomy. However, despite abundance declines, high between-assemblage variability at deeper depths allowed biodiversity-supporting potential to be maintained, highlighting their importance as coastal refugia.
- 6 April 2021
- Published by Taylor & Francis in Journal of the American Statistical Association
- Vol. 117 (540) , 1763-1774
On the Hauck–Donner Effect in Wald Tests: Detection, Tipping Points, and Parameter Space Characterization
The Wald test remains ubiquitous in statistical practice despite shortcomings such as its inaccuracy in small samples and lack of invariance under reparameterization. This paper develops on another but lesser-known shortcoming called the Hauck–Donner effect (HDE) whereby a Wald test statistic is no longer monotone increasing as a function of increasing distance between the parameter estimate and the null value. Resulting in an upward biased p-value and loss of power, the aberration can lead to very damaging consequences such as in variable selection. The HDE afflicts many types of regression models and corresponds to estimates near the boundary of the parameter space. This article presents several new results, and its main contributions are to (i) propose a very general test for detecting the HDE in the class of vector generalized linear models (VGLMs), regardless of the underlying cause; (ii) fundamentally characterize the HDE by pairwise ratios of Wald and Rao score and likelihood ratio test statistics for 1-parameter distributions with large samples; (iii) show that the parameter space may be partitioned into an interior encased by at least 5 HDE severity measures (faint, weak, moderate, strong, extreme); (iv) prove that a necessary condition for the HDE in a 2 by 2 table is a log odds ratio of at least 2; (v) give some practical guidelines about HDE-free hypothesis testing. Overall, practical post-fit tests can now be conducted potentially to any model estimated by iteratively reweighted least squares, especially the GLM and VGLM classes, the latter which encompasses many popular regression models.
- 6 April 2021
- Published by Taylor & Francis in Journal of the American Statistical Association
- Vol. 117 (540) , 1763-1774
On the Hauck–Donner Effect in Wald Tests: Detection, Tipping Points, and Parameter Space Characterization
The Wald test remains ubiquitous in statistical practice despite shortcomings such as its inaccuracy in small samples and lack of invariance under reparameterization. This paper develops on another but lesser-known shortcoming called the Hauck–Donner effect (HDE) whereby a Wald test statistic is no longer monotone increasing as a function of increasing distance between the parameter estimate and the null value. Resulting in an upward biased p-value and loss of power, the aberration can lead to very damaging consequences such as in variable selection. The HDE afflicts many types of regression models and corresponds to estimates near the boundary of the parameter space. This article presents several new results, and its main contributions are to (i) propose a very general test for detecting the HDE in the class of vector generalized linear models (VGLMs), regardless of the underlying cause; (ii) fundamentally characterize the HDE by pairwise ratios of Wald and Rao score and likelihood ratio test statistics for 1-parameter distributions with large samples; (iii) show that the parameter space may be partitioned into an interior encased by at least 5 HDE severity measures (faint, weak, moderate, strong, extreme); (iv) prove that a necessary condition for the HDE in a 2 by 2 table is a log odds ratio of at least 2; (v) give some practical guidelines about HDE-free hypothesis testing. Overall, practical post-fit tests can now be conducted potentially to any model estimated by iteratively reweighted least squares, especially the GLM and VGLM classes, the latter which encompasses many popular regression models.
- 1 March 2020
- Published by Wiley in Australian & New Zealand Journal of Statistics
- Vol. 62 (1) , 116-131
The VGAM package for negative binomial regression
Negative binomial (NB) regression is the most common full‐likelihood method for analysing count data exhibiting overdispersion with respect to the Poisson distribution. Usually most practitioners are content to fit one of two NB variants, however other important variants exist. It is demonstrated here that theVGAM Rpackage can fit them all under a common statistical framework founded upon a generalised linear and additive model approach. Additionally, other modifications such as zero‐altered (hurdle), zero‐truncated and zero‐inflated NB distributions are naturally handled. Rootograms are also available for graphically checking the goodness of fit. Two data sets and some recently added features of theVGAMpackage are used here for illustration.
- 1 March 2020
- Published by Wiley in Australian & New Zealand Journal of Statistics
- Vol. 62 (1) , 116-131
The VGAM package for negative binomial regression
Negative binomial (NB) regression is the most common full‐likelihood method for analysing count data exhibiting overdispersion with respect to the Poisson distribution. Usually most practitioners are content to fit one of two NB variants, however other important variants exist. It is demonstrated here that theVGAM Rpackage can fit them all under a common statistical framework founded upon a generalised linear and additive model approach. Additionally, other modifications such as zero‐altered (hurdle), zero‐truncated and zero‐inflated NB distributions are naturally handled. Rootograms are also available for graphically checking the goodness of fit. Two data sets and some recently added features of theVGAMpackage are used here for illustration.
- Open Access
- 7 May 2019
- Published by Wiley in Journal of Probability and Statistics
- Vol. 2019, 1-11
New Link Functions for Distribution–Specific Quantile Regression Based on Vector Generalized Linear and Additive Models
In the usual quantile regression setting, the distribution of the response given the explanatory variables is unspecified. In this work, the distribution is specified and we introduce new link functions to directly model specified quantiles of seven 1–parameter continuous distributions. Using the vector generalized linear and additive model (VGLM/VGAM) framework, we transform certain prespecified quantiles to become linear or additive predictors. Our parametric quantile regression approach adopts VGLMs/VGAMs because they can handle multiple linear predictors and encompass many distributions beyond the exponential family. Coupled with the ability to fit smoothers, the underlying strong assumption of the distribution can be relaxed so as to offer a semiparametric–type analysis. By allowing multiple linear and additive predictors simultaneously, the quantile crossing problem can be avoided by enforcing parallelism constraint matrices. This article gives details of a software implementation called the VGAMextra package for R. Both the data and recently developed software used in this paper are freely downloadable from the internet.
- 7 May 2019
- Published by Wiley in Journal of Probability and Statistics
- Vol. 2019, 1-11
Open AccessNew Link Functions for Distribution–Specific Quantile Regression Based on Vector Generalized Linear and Additive Models
In the usual quantile regression setting, the distribution of the response given the explanatory variables is unspecified. In this work, the distribution is specified and we introduce new link functions to directly model specified quantiles of seven 1–parameter continuous distributions. Using the vector generalized linear and additive model (VGLM/VGAM) framework, we transform certain prespecified quantiles to become linear or additive predictors. Our parametric quantile regression approach adopts VGLMs/VGAMs because they can handle multiple linear predictors and encompass many distributions beyond the exponential family. Coupled with the ability to fit smoothers, the underlying strong assumption of the distribution can be relaxed so as to offer a semiparametric–type analysis. By allowing multiple linear and additive predictors simultaneously, the quantile crossing problem can be avoided by enforcing parallelism constraint matrices. This article gives details of a software implementation called the VGAMextra package for R. Both the data and recently developed software used in this paper are freely downloadable from the internet.
- 1 October 2016
- Published by Taylor & Francis in Journal of the American Statistical Association
- Vol. 111 (516) , 1565-1568
Comment
- 1 October 2016
- Published by Taylor & Francis in Journal of the American Statistical Association
- Vol. 111 (516) , 1565-1568
Comment