Guanyu Chen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Guanyu Chen has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
Amery D. Wu
3 shared publicationsChiyu Zhang
2 shared publicationsYan Liu
2 shared publications- Yan Liu
Carleton University
2 shared publications Audrey Béliveau
Department of Statistics and Actuarial Science, University of Waterloo
1 shared publicationBin Tan
University of Alberta
1 shared publication
Latest Publications
- Open Access
- 1 October 2026
- Published by Springer Nature in Large-scale Assessments in Education
- Vol. 14 (1) , 50
Educational productivity in high-performing systems: a comparison of Hong Kong and Singapore on PIRLS 2021
Background: Reading achievement is a critical educational outcome with long-term implications for students’ academic and social development. While both Hong Kong and Singapore consistently perform at high levels in international assessments, most existing research has examined these systems independently. Few studies have directly compared how student, family, and school factors operate across these two top-performing Asian systems. This study addresses that gap by exploring contextual factors related to reading achievement within Walberg’s theory of educational productivity. By identifying both shared and distinctive patterns, the study aims to inform local policy efforts and provide insights for other systems seeking to improve reading outcomes. Methods: Data were drawn from the Progress in International Reading Literacy Study 2021. Hierarchical linear modeling (HLM) was conducted separately for Hong Kong and Singapore to examine the contributions of variables across four domains: student characteristics, student attitudes and behaviors, family background, and school context. HLM accounts for the nested structure of students within schools, allowing for accurate estimation at both the student and school levels. Results: Predictors of reading achievement were identified across all domains. Absenteeism was associated with lower reading scores in both systems, while higher reading confidence and more frequent reading outside school were linked to better performance. The comparison between Hong Kong and Singapore showed different patterns in how family and school factors related to literacy outcomes. A stronger socioeconomic background was associated with better reading performance, with a larger estimated coefficient in Singapore. In Singapore, parental reading habits, parental perceptions of schooling, and schools’ emphasis on academic success were all positively related to reading achievement. In contrast, these relationships were weak or nonsignificant in Hong Kong. Conclusions: This study examines reading achievement through a multilevel lens across education systems. Both systems face challenges such as absenteeism and require stronger support structures. It remains important for both systems to continue narrowing performance gaps between privileged and underprivileged students by promoting deeper equity in education and encouraging more coordinated family–school collaboration to enhance reading achievement.
- 1 October 2026
- Published by Springer Nature in Large-scale Assessments in Education
- Vol. 14 (1) , 50
Open AccessEducational productivity in high-performing systems: a comparison of Hong Kong and Singapore on PIRLS 2021
Background: Reading achievement is a critical educational outcome with long-term implications for students’ academic and social development. While both Hong Kong and Singapore consistently perform at high levels in international assessments, most existing research has examined these systems independently. Few studies have directly compared how student, family, and school factors operate across these two top-performing Asian systems. This study addresses that gap by exploring contextual factors related to reading achievement within Walberg’s theory of educational productivity. By identifying both shared and distinctive patterns, the study aims to inform local policy efforts and provide insights for other systems seeking to improve reading outcomes. Methods: Data were drawn from the Progress in International Reading Literacy Study 2021. Hierarchical linear modeling (HLM) was conducted separately for Hong Kong and Singapore to examine the contributions of variables across four domains: student characteristics, student attitudes and behaviors, family background, and school context. HLM accounts for the nested structure of students within schools, allowing for accurate estimation at both the student and school levels. Results: Predictors of reading achievement were identified across all domains. Absenteeism was associated with lower reading scores in both systems, while higher reading confidence and more frequent reading outside school were linked to better performance. The comparison between Hong Kong and Singapore showed different patterns in how family and school factors related to literacy outcomes. A stronger socioeconomic background was associated with better reading performance, with a larger estimated coefficient in Singapore. In Singapore, parental reading habits, parental perceptions of schooling, and schools’ emphasis on academic success were all positively related to reading achievement. In contrast, these relationships were weak or nonsignificant in Hong Kong. Conclusions: This study examines reading achievement through a multilevel lens across education systems. Both systems face challenges such as absenteeism and require stronger support structures. It remains important for both systems to continue narrowing performance gaps between privileged and underprivileged students by promoting deeper equity in education and encouraging more coordinated family–school collaboration to enhance reading achievement.
- Open Access
- 23 July 2026
- Published by Center for Open Science in PsyArXiv
Assessing Fairness in AI-Assisted Writing Scoring: Two Indices for Quantifying the Magnitude of Group Differences
Fairness evaluation in automated essay scoring (AES) requires more than determining whether group differences exist. It also requires evidence about how large those differences are, which group they favor, and whether they warrant further attention. This study introduces two model-level indices, Conditional Mean Disparity (CMD) and Conditional Odds Disparity (COD), for evaluating group differences in predicted writing scores after accounting for human-rated proficiency. CMD quantifies disparity on the predicted score scale, whereas COD quantifies disparity using the relative odds of receiving higher predicted scores. The indices were demonstrated using writing samples from the PERSUADE 2.0 corpus and four AES systems representing traditional statistical modeling, deep learning, and prompt-based large language model scoring. The findings showed that CMD and COD provided interpretable estimates of disparity magnitude across systems and captured information that was not reflected in scoring performance alone. By emphasizing the size and direction of observed group differences, the proposed indices can support fairness evaluation, model comparison, and decisions about whether further examination is needed.
- 23 July 2026
- Published by Center for Open Science in PsyArXiv
Open AccessAssessing Fairness in AI-Assisted Writing Scoring: Two Indices for Quantifying the Magnitude of Group Differences
Fairness evaluation in automated essay scoring (AES) requires more than determining whether group differences exist. It also requires evidence about how large those differences are, which group they favor, and whether they warrant further attention. This study introduces two model-level indices, Conditional Mean Disparity (CMD) and Conditional Odds Disparity (COD), for evaluating group differences in predicted writing scores after accounting for human-rated proficiency. CMD quantifies disparity on the predicted score scale, whereas COD quantifies disparity using the relative odds of receiving higher predicted scores. The indices were demonstrated using writing samples from the PERSUADE 2.0 corpus and four AES systems representing traditional statistical modeling, deep learning, and prompt-based large language model scoring. The findings showed that CMD and COD provided interpretable estimates of disparity magnitude across systems and captured information that was not reflected in scoring performance alone. By emphasizing the size and direction of observed group differences, the proposed indices can support fairness evaluation, model comparison, and decisions about whether further examination is needed.
- Open Access
- 19 May 2026
- Published by Elsevier in Assessing Writing
- Vol. 69
Assessing fairness in AI-assisted writing scoring: Developing fairness measures to detect predictive bias in automated essay scoring
- 19 May 2026
- Published by Elsevier in Assessing Writing
- Vol. 69
Open AccessAssessing fairness in AI-assisted writing scoring: Developing fairness measures to detect predictive bias in automated essay scoring
- Open Access
- 12 May 2026
- Published by Center for Open Science in PsyArXiv
Standardizing Predictors for Computing Effect Sizes in Multilevel Modeling with Random Slopes
Reporting effect sizes has become standard practice in the social and behavioral sciences, and multilevel modeling is widely used for analyzing hierarchically structured data. However, calculating effect sizes for multilevel modeling with random slopes remains challenging because the variance of the random effects depends on the values of the predictor, resulting in heteroscedasticity. Existing solutions address this issue by treating the random effects variance as a mixture of individual variances, which can be computationally complex and difficult to interpret. This study addresses this gap by proposing a simple and practical alternative: standardizing predictors associated with random slopes to yield a straightforward variance decomposition that facilitates effect size calculation. We evaluate this approach using both an empirical data example and a simulation study. Results show that the predictor standardization approach is easy to implement and produces effect size estimates comparable to those obtained using the mixture distribution approach. This work provides a practical solution for effect size reporting in multilevel modeling with random slopes and supports broader adoption of effect size measures in applied multilevel research.
- 12 May 2026
- Published by Center for Open Science in PsyArXiv
Open AccessStandardizing Predictors for Computing Effect Sizes in Multilevel Modeling with Random Slopes
Reporting effect sizes has become standard practice in the social and behavioral sciences, and multilevel modeling is widely used for analyzing hierarchically structured data. However, calculating effect sizes for multilevel modeling with random slopes remains challenging because the variance of the random effects depends on the values of the predictor, resulting in heteroscedasticity. Existing solutions address this issue by treating the random effects variance as a mixture of individual variances, which can be computationally complex and difficult to interpret. This study addresses this gap by proposing a simple and practical alternative: standardizing predictors associated with random slopes to yield a straightforward variance decomposition that facilitates effect size calculation. We evaluate this approach using both an empirical data example and a simulation study. Results show that the predictor standardization approach is easy to implement and produces effect size estimates comparable to those obtained using the mixture distribution approach. This work provides a practical solution for effect size reporting in multilevel modeling with random slopes and supports broader adoption of effect size measures in applied multilevel research.
- Open Access
- 1 January 2026
- Published by Canadian Science Publishing in FACETS
- Vol. 11, 1-22
Tracking learning loss in British Columbia: evidence of post-pandemic declines in foundational numeracy and literacy skills (2017–2023)
Potential learning loss associated with the COVID-19 pandemic was examined by analyzing British Columbia’s Foundational Skills Assessment data for students in grades 4 and 7 from 2017 to 2023. Drawing on over one million students’ data, results reveal a consistent downward trend in numeracy and literacy (i.e., reading comprehension) scores, with 2022/2023 marking the lowest performance across the six academic-year span. Gender disparities widened in numeracy during the post-pandemic period, with boys scoring significantly higher than girls, particularly in 2022/2023. In literacy, girls consistently outperformed boys, though the size of the gender gap varied over time and remained statistically significant. A persistent rural–urban gap was observed in both domains, with urban students consistently outperforming their rural counterparts. Notably, both groups experienced significant declines in literacy post-pandemic, despite slight improvements from 2021/2022 to 2022/2023. These findings suggest that the educational disruptions caused by the pandemic have had lasting effects on foundational academic skills, with differential impacts across gender and geographic location. Continued monitoring and targeted interventions are needed to address these disparities and support students’ academic recovery.
- 1 January 2026
- Published by Canadian Science Publishing in FACETS
- Vol. 11, 1-22
Open AccessTracking learning loss in British Columbia: evidence of post-pandemic declines in foundational numeracy and literacy skills (2017–2023)
Potential learning loss associated with the COVID-19 pandemic was examined by analyzing British Columbia’s Foundational Skills Assessment data for students in grades 4 and 7 from 2017 to 2023. Drawing on over one million students’ data, results reveal a consistent downward trend in numeracy and literacy (i.e., reading comprehension) scores, with 2022/2023 marking the lowest performance across the six academic-year span. Gender disparities widened in numeracy during the post-pandemic period, with boys scoring significantly higher than girls, particularly in 2022/2023. In literacy, girls consistently outperformed boys, though the size of the gender gap varied over time and remained statistically significant. A persistent rural–urban gap was observed in both domains, with urban students consistently outperforming their rural counterparts. Notably, both groups experienced significant declines in literacy post-pandemic, despite slight improvements from 2021/2022 to 2022/2023. These findings suggest that the educational disruptions caused by the pandemic have had lasting effects on foundational academic skills, with differential impacts across gender and geographic location. Continued monitoring and targeted interventions are needed to address these disparities and support students’ academic recovery.
- Open Access
- 1 January 2025
- Published by Wiley in Depression and Anxiety
- Vol. 2025 (1) , 6663877
Psychological Distress in Childbearing Persons During the COVID‐19 Pandemic: A Multi‐Trajectory Study of Anger, Anxiety, and Depression
Psychological distress can manifest as depression, anxiety, and anger in the perinatal period. These conditions are often comorbid yet studied in isolation. A full understanding of perinatal psychopathology requires the spectrum of common psychological distress to be studied concurrently to better understand interconnected symptoms. A transdiagnostic approach provides valuable insights into how symptoms interact and cumulatively affect mental health, which can inform more effective screening and treatment strategies. This, in turn, can improve outcomes for birthing parents experiencing psychological distress. We undertook group‐based multi‐trajectory modeling (GBMTM) to uncover the patterns of affective disorders (anger, anxiety, and depression) over three‐time points (pregnancy, 3‐, and 12‐months postpartum (mPP)) in a large longitudinal cohort of persons who gave birth during the COVID‐19 pandemic ( n = 2145). We identified five trajectory groups: high‐stable (11.3%), postpartum‐increase (16.0%), postpartum‐decrease (21.5%), low‐stable (37.9%), and minimal stable (13.2%) symptoms of anger, anxiety, and depression. Multinomial regression revealed that lower levels of sleep disturbance, less financial hardship, and lower intolerance of uncertainty predicted postpartum decreases in psychological distress compared with the high stable group. Higher levels of sleep disturbance, greater financial hardship, lower level of social support, and greater intolerance of uncertainty predicted postpartum increases in psychological distress compared with the low‐stable and minimal‐stable groups. Screening for psychological distress symptoms (i.e., anger, anxiety, and depression), paired with access to evidence‐based management for those who screen positive, is warranted during the first postpartum year to reduce the harmful effects of unmanaged distress on families.
- 1 January 2025
- Published by Wiley in Depression and Anxiety
- Vol. 2025 (1) , 6663877
Open AccessPsychological Distress in Childbearing Persons During the COVID‐19 Pandemic: A Multi‐Trajectory Study of Anger, Anxiety, and Depression
Psychological distress can manifest as depression, anxiety, and anger in the perinatal period. These conditions are often comorbid yet studied in isolation. A full understanding of perinatal psychopathology requires the spectrum of common psychological distress to be studied concurrently to better understand interconnected symptoms. A transdiagnostic approach provides valuable insights into how symptoms interact and cumulatively affect mental health, which can inform more effective screening and treatment strategies. This, in turn, can improve outcomes for birthing parents experiencing psychological distress. We undertook group‐based multi‐trajectory modeling (GBMTM) to uncover the patterns of affective disorders (anger, anxiety, and depression) over three‐time points (pregnancy, 3‐, and 12‐months postpartum (mPP)) in a large longitudinal cohort of persons who gave birth during the COVID‐19 pandemic ( n = 2145). We identified five trajectory groups: high‐stable (11.3%), postpartum‐increase (16.0%), postpartum‐decrease (21.5%), low‐stable (37.9%), and minimal stable (13.2%) symptoms of anger, anxiety, and depression. Multinomial regression revealed that lower levels of sleep disturbance, less financial hardship, and lower intolerance of uncertainty predicted postpartum decreases in psychological distress compared with the high stable group. Higher levels of sleep disturbance, greater financial hardship, lower level of social support, and greater intolerance of uncertainty predicted postpartum increases in psychological distress compared with the low‐stable and minimal‐stable groups. Screening for psychological distress symptoms (i.e., anger, anxiety, and depression), paired with access to evidence‐based management for those who screen positive, is warranted during the first postpartum year to reduce the harmful effects of unmanaged distress on families.
- 29 November 2024
- Published by Taylor & Francis in Basic and Applied Social Psychology
- Vol. 47 (2) , 105-119
A Bibliometric Review of Natural Language Processing Applications in Psychology from 1991 to 2023
- 29 November 2024
- Published by Taylor & Francis in Basic and Applied Social Psychology
- Vol. 47 (2) , 105-119
A Bibliometric Review of Natural Language Processing Applications in Psychology from 1991 to 2023
- Open Access
- 11 October 2024
- Published by Wiley in British Journal of Developmental Psychology
- Vol. 43 (1) , 82-97
Parental control and adolescent social anxiety: A focus on emotional regulation strategies and socioeconomic influences in China
An individual's social adaptation, which is influenced by both internalizing and externalizing factors, depends on social anxiety. We proposed that the connection between parental psychological control and social anxiety among middle school students was mediated by negative emotion response‐focused strategies and moderated by socioeconomic status (SES). We collected data from 1343 Chinese students from 12 to 16 years old (M = 13.3, SD = 0.96) by applying the questionnaire of parental psychological control scale, social anxiety scale and negative emotion response‐focused strategies scale. The findings demonstrated that parental psychological control, including father and mother psychological control, was positively associated with social anxiety and that the relationship between parental psychological control and children's social anxiety was mediated by negative emotion response‐focused strategies. Meanwhile, SES played a moderating role in the relationship between children's negative emotion response‐focused strategies and social anxiety. The findings also suggest that we should pay more attention to the psychological status and social interaction of children with lower SES.
- 11 October 2024
- Published by Wiley in British Journal of Developmental Psychology
- Vol. 43 (1) , 82-97
Open AccessParental control and adolescent social anxiety: A focus on emotional regulation strategies and socioeconomic influences in China
An individual's social adaptation, which is influenced by both internalizing and externalizing factors, depends on social anxiety. We proposed that the connection between parental psychological control and social anxiety among middle school students was mediated by negative emotion response‐focused strategies and moderated by socioeconomic status (SES). We collected data from 1343 Chinese students from 12 to 16 years old (M = 13.3, SD = 0.96) by applying the questionnaire of parental psychological control scale, social anxiety scale and negative emotion response‐focused strategies scale. The findings demonstrated that parental psychological control, including father and mother psychological control, was positively associated with social anxiety and that the relationship between parental psychological control and children's social anxiety was mediated by negative emotion response‐focused strategies. Meanwhile, SES played a moderating role in the relationship between children's negative emotion response‐focused strategies and social anxiety. The findings also suggest that we should pay more attention to the psychological status and social interaction of children with lower SES.
- Open Access
- 23 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304109
Understanding the log file data from educational and psychological computer-based testing: A scoping review protocol
With the advancement of computer-based testing, log file data has drawn considerable attention from researchers. Although emerging studies have begun to explore log file data, there is a gap in the exploitation of log file data for capturing and understanding participants’ cognitive processes. The debate on how to maximize insights from log file data has not yet reached a consensus. Therefore, we present this protocol for a scoping review that aims to characterize the application of log file data in current publications, including the data pre-processing techniques, analytical methodologies, and theoretical frameworks used by researchers. This review will also aim to illuminate how log file data can enhance psychological and educational assessments. Our findings will highlight the opportunities and challenges presented by log file data as an emerging and essential source of evidence for future advancements in psychological and educational assessment.
- 23 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304109
Open AccessUnderstanding the log file data from educational and psychological computer-based testing: A scoping review protocol
With the advancement of computer-based testing, log file data has drawn considerable attention from researchers. Although emerging studies have begun to explore log file data, there is a gap in the exploitation of log file data for capturing and understanding participants’ cognitive processes. The debate on how to maximize insights from log file data has not yet reached a consensus. Therefore, we present this protocol for a scoping review that aims to characterize the application of log file data in current publications, including the data pre-processing techniques, analytical methodologies, and theoretical frameworks used by researchers. This review will also aim to illuminate how log file data can enhance psychological and educational assessments. Our findings will highlight the opportunities and challenges presented by log file data as an emerging and essential source of evidence for future advancements in psychological and educational assessment.
- 18 October 2022
- Published by Taylor & Francis in Multivariate Behavioral Research
- Vol. 58 (4) , 706-722
A Gentle Introduction to Bayesian Network Meta-Analysis Using an Automated R Package
Network meta-analysis is an extension of standard meta-analysis. It allows researchers to build a network of evidence to compare multiple interventions that may have not been compared directly in existing publications. With a Bayesian approach, network meta-analysis can be used to obtain a posterior probability distribution of all the relative treatment effects, which allows for the estimation of relative treatment effects to quantify the uncertainty of parameter estimates, and to rank all the treatments in the network. Ranking treatments using both direct and indirect evidence can provide guidance to policy makers and clinicians for making decisions. The purpose of this paper is to introduce fundamental concepts of Bayesian network meta-analysis (BNMA) to researchers in psychology and social sciences. We discuss several essential concepts of BNMA, including the assumptions of homogeneity and consistency, the fixed and random effects models, prior specification, and model fit evaluation strategies, while pointing out some issues and areas where researchers should use caution in the application of BNMA. Additionally, using an automated R package, we provide a step-by-step demonstration on how to conduct and report the findings of BNMA with a real dataset of psychological interventions extracted from PubMed.
- 18 October 2022
- Published by Taylor & Francis in Multivariate Behavioral Research
- Vol. 58 (4) , 706-722
A Gentle Introduction to Bayesian Network Meta-Analysis Using an Automated R Package
Network meta-analysis is an extension of standard meta-analysis. It allows researchers to build a network of evidence to compare multiple interventions that may have not been compared directly in existing publications. With a Bayesian approach, network meta-analysis can be used to obtain a posterior probability distribution of all the relative treatment effects, which allows for the estimation of relative treatment effects to quantify the uncertainty of parameter estimates, and to rank all the treatments in the network. Ranking treatments using both direct and indirect evidence can provide guidance to policy makers and clinicians for making decisions. The purpose of this paper is to introduce fundamental concepts of Bayesian network meta-analysis (BNMA) to researchers in psychology and social sciences. We discuss several essential concepts of BNMA, including the assumptions of homogeneity and consistency, the fixed and random effects models, prior specification, and model fit evaluation strategies, while pointing out some issues and areas where researchers should use caution in the application of BNMA. Additionally, using an automated R package, we provide a step-by-step demonstration on how to conduct and report the findings of BNMA with a real dataset of psychological interventions extracted from PubMed.