Research
Recent Research Themes
1. Developments in Generalized Structured Component Analysis Structural Equation Modeling (GSCA-SEM)
Structural Equation Modeling (SEM) is a popular multivariate statistical tool for specifying and examining the relationships between observed variables and constructs in a wide range of sciences.
The Lab has developed Generalized Structured Component Analysis Structural Equation Modeling (GSCA-SEM) for estimating various models where constructs are represented by factors and/or components. It has been actively extending and refining GSCA-SEM to improve its data-analytic flexibility and generality. The Lab also has implemented GSCA-SEM into free software programs (e.g., GSCA Pro).
2. Development of Data Integration Methods
Data integration represents combining (often large) data residing in multiple and varied sources (e.g., subjects, stimuli, areas, etc.) to provide a unified and simplified view of the data. This becomes important in various scientific domains.
The Lab has developed flexible tools for data integration. For instance, it extended multiple-set canonical correlation analysis to the analysis of functional data that are generated from smooth curves varying over a continuum. This method was applied to fMRI data to identify networks of neural activity which were commonly activated across several subjects performing the same working memory task (Hwang et al., 2012). Moreover, the Lab combined principal component analysis with multiple-set canonical correlation analysis in a unified framework to integrate multiple sources of data (Choi et al., 2017; Hwang et al., 2013).
3. Effective Connectivity Analysis of Neuroimaging Data
It is well recognized that a particular behavioural or cognitive task is associated with neural networks of multiple brain regions, rather than isolated brain regions. Connectivity analysis can describe relationships between brain regions. There are two different approaches to connectivity analysis – functional vs. effective connectivity. In particular, effective connectivity analysis focuses on directional relationships among a set of interconnected brain regions that are selected based on a hypothesis or prior knowledge about their importance in completing a task. This approach may be used to better explain functional integration within a distributed neural system, allowing quantifications and stronger inferences of directed connections of different brain region activities.
The Lab has developed structural equation models for effective connectivity analysis (e.g., Cho et al., 2014; Jung et al., 2012, 2016; Zhou et al., 2016).
From Cho et al. (2014)
4. Knowledge-Based Multivariate Analysis of Genetic Data
Over the decade, a substantial amount of information has been accumulated for researchers to specify which candidate genotypes, such as single nucleotide polymorphisms (SNPs), are linked to which genes, as well as which genes might be associated with which traits or phenotypes in specific clinical populations.
The Lab has collaborated with biostatisticians to develop statistical methods for testing the associations between genotypes and traits in a biologically meaningful and confirmatory manner based on previous knowledge or hypotheses (e.g., Lee et al., 2016, 2018; Romdhani et al., 2015). The Lab is currently extending these methods to take into account biologically important issues (e.g., polygenicity, leiotropy, epistatisis, etc.).
5. Imaging Genetics Multivariate Modeling for Examining Gene-Brain-Behavioural/Cognitive Relationships
Imaging genetics is a rapidly emerging field that examines genetic influences on altered brain activities associated with behavioural or cognitive variation.
The Lab has developed a general statistical method, named Imaging Genetics Generalized Structured Component Analysis (IG-GSCA; Hwang et al., 2021), for investigating associations among genetic, brain, and behavioural/cognitive phenotypes, while taking into account biological complexities (e.g., genetic networks, gene-gene interactions, gene-environment interactions, etc.) and methodological issues (e.g., multicollinearity). The Lab is currently involved in various technical and empirical developments of IG-GSCA.
From Tost et al. (2012)
6. Theoretical Development in Functional Data Analysis
Functional data represent data collected in the form of curves, surfaces, images, or anything else varying over a continuum. The continuum can be time, spatial position, wavelength, probability, and so forth. Owing to the emergence of various novel measurement tools such as eye-trackers, motion capture devices, and functional neuroimaging modalities, functional data become ubiquitous in psychology and other social sciences. Functional data analysis is a fast-emerging domain of statistics in which such functional data are taken as the basic elements of analysis. The Lab has contributed to theoretical development in this exciting domain (e.g., Choi et al., 2017, 2018; Hwang et al., 2012, 2015; Suk & Hwang, 2016; Tan et al., 2013, 2015).