[05/25/2024] Psychometric Network Analysis
Last updated
Last updated
Investigating the performance of exploratory graph analysis and traditional techniques to identify the number of latent factors: A simulation and tutorial
, , , , , , ,
Abstract
Exploratory graph analysis (EGA) is a new technique that was recently proposed within the framework of network psychometrics to estimate the number of factors underlying multivariate data. Unlike other methods, EGA produces a visual guide––network plot––that not only indicates the number of dimensions to retain, but also which items cluster together and their level of association. Although previous studies have found EGA to be superior to traditional methods, they are limited in the conditions considered. These issues are here addressed through an extensive simulation study that incorporates a wide range of plausible structures that may be found in practice, including continuous and dichotomous data, and unidimensional and multidimensional structures. Additionally, two new EGA techniques are presented, one that extends EGA to also deal with unidimensional structures, and the other based on the triangulated maximally filtered graph approach (EGAtmfg). Both EGA techniques are compared with five widely used factor analytic techniques. Overall, EGA and EGAtmfg are found to perform as well as the most accurate traditional method, parallel analysis, and to produce the best large-sample properties of all the methods evaluated. To facilitate the use and application of EGA, we present a straightforward R tutorial on how to apply and interpret EGA, using scores from a well-known psychological instrument: the Marlowe-Crowne Social Desirability Scale
Scale Development via Network Analysis: A Comprehensive and Concise Measure of Openness to Experience
, , , and
Abstract
Abstract
Psychometric network analysis is an emerging tool to investigate the structure of psychological and psychopathological constructs. To date, most of the psychometric network literature has emphasized the measurement of constructs (e.g., dimensional structure); however, this represents only one aspect of psychometrics. In the present study, we explored whether network analysis could be used as a tool for scale development. To do so, we used a previously published dataset (N = 794) of four Openness to Experience inventories to clarify the facet structure of the construct and identify the conceptual coverage of each inventory. In short, 10 facets and 3 aspects (i.e., meso-facets) were identified but no single inventory adequately covered all facets or aspects. Therefore, we used network analysis, including two novel network measures (community closeness centrality and network coverage), to develop a short measure that comprehensively measured all facets and aspects of the construct. We then compared the network-derived short form to short forms that were developed using classical test theory (CTT) and item response theory (IRT). The network-derived short form demonstrated comparable reliability to the CTT- and IRT-based short forms but had better coverage of the conceptual space (defined by the four inventories). Finally, we validated the network-derived short form by comparing its correlations with outcome measures (personality and political conservatism) to that of the four-inventory item pool. We found that the network approach is a promising psychometric tool for scale development and we discuss its implications for future applications.Show less
A Roadmap for sale Development using Network Analysis
Exploratory graph analysis.
Community closeness centrality
Hybrid centrality
Unique variable analysis: A network psychometrics method to detect local dependence.
Abstract
The local independence assumption states that variables are unrelated after conditioning on a latent variable. Common problems that arise from violations of this assumption include model misspecification, biased model parameters, and inaccurate estimates of internal structure. These problems are not limited to latent variable models but also apply to network psychometrics. This paper proposes a novel network psychometric approach to detect locally dependent pairs of variables using network modeling and a graph theory measure called weighted topological overlap (wTO). Using simulation, this approach is compared to contemporary local dependence detection methods such as exploratory structural equation modeling with standardized expected parameter change and a recently developed approach using partial correlations and a resampling procedure. Different approaches to determine local dependence using statistical significance and cutoff values are also compared. Continuous, polytomous (5-point Likert scale), and dichotomous (binary) data were generated with skew across a variety of conditions. Our results indicate that cutoff values work better than significance approaches. Overall, the network psychometrics approaches using wTO with graphical least absolute shrinkage and selector operator with extended Bayesian information criterion and wTO with Bayesian Gaussian graphical model were the best performing local dependence detection methods overall.
Semantic network analysis (SemNA): A tutorial on preprocessing, estimating, and analyzing semantic networks
Abstract
To date, the application of semantic network methodologies to study cognitive processes in psychological phenomena has been limited in scope. One barrier to broader application is the lack of resources for researchers unfamiliar with the approach. Another barrier, for both the unfamiliar and knowledgeable researcher, is the tedious and laborious preprocessing of semantic data. We aim to minimize these barriers by offering a comprehensive semantic network analysis pipeline (preprocessing, estimating, and analyzing networks), and an associated R tutorial that uses a suite of R packages to accommodate the pipeline. Two of these packages, SemNetDictionaries and SemNetCleaner, promote an efficient, reproducible, and transparent approach to preprocessing linguistic data. The third package, SemNeT, provides methods and measures for estimating and statistically comparing semantic networks via a point-and-click graphical user interface. Using real-world data, we present a start-to-finish pipeline from raw data to semantic network analysis results. This article aims to provide resources for researchers, both the unfamiliar and knowledgeable, that reduce some of the barriers for conducting semantic network analysis.
Factor or Network Model? Predictions FromNeural Networks
Abstract
Abstract.The nature of associations between variables is important forconstructing theory about psychological phenomena. In the last decade,this topic has received renewed interest with the introduction of psycho-metric network models. In psychology, network models are often con-trasted with latent variable (e.g., factor) models. Recent research hasshown that differences between the two tend to be more substantivethan statistical. One recently developed algorithm called theLoadingsComparison Test(LCT) was developed to predict whether data weregenerated from a factor or small-world network model. A significant limi-tation of the current LCT implementation is that it’s based on heuristicsthat were derived from descriptive statistics. In the present study, weused artificial neural networks to replace these heuristics and develop amore robust and generalizable algorithm. We performed a Monte Carlosimulation study that compared neural networks to the original LCTalgorithm as well as logistic regression models that were trained on thesame data. We found that the neural networks performed as well as orbetter than both methods for predicting whether data were generatedfrom a factor, small-world network, or random network model. Althoughthe neural networks were trained on small-world networks, we show thatthey can reliably predict the data-generating model of random networks,demonstrating generalizability beyond the trained data. We echo the callfor more formal theories about the relations between variables and dis-cuss the role of the LCT in this process.
Comparing community detection algorithms in psychometric networks: A Monte Carlo simulation
Abstract
Identifying the correct number of factors in multivariate data is fundamental to psychological measurement. Factor analysis has a long tradition in the field, but it has been challenged recently by exploratory graph analysis (EGA), an approach based on network psychometrics. EGA first estimates a network and then applies the Walktrap community detection algorithm. Simulation studies have demonstrated that EGA has comparable or better accuracy for recovering the same number of communities as there are factors in the simulated data than factor analytic methods. Despite EGA’s effectiveness, there has yet to be an investigation into whether other sparsity induction methods or community detection algorithms could achieve equivalent or better performance. Furthermore, unidimensional structures are fundamental to psychological measurement yet they have been sparsely studied in simulations using community detection algorithms. In the present study, we performed a Monte Carlo simulation using the zero-order correlation matrix, GLASSO, and two variants of a non-regularized partial correlation sparsity induction methods with several community detection algorithms. We examined the performance of these method–algorithm combinations in both continuous and polytomous data across a variety of conditions. The results indicate that the Fast-greedy, Louvain, and Walktrap algorithms paired with the GLASSO method were consistently among the most accurate and least-biased overall.
Adjusted Network Loadings for Dynamic Exploratory Graph Analysis
Laura Jamison, Hudson Golino, Alexander P Christensen
Abstract
The possibility of computing network loadings (analogous to factor loadings) in the network psychometrics framework (specifically, Exploratory Graph Analysis (EGA)) was recently made possible. In the cross-sectional case, it has been demonstrated that network loadings have a linear, symmetric relationship to their corresponding true population loadings. It was demonstrated that Dynamic EGA (DynEGA) has the ability to produce network scores and network loadings with the same calculations as in the cross-sectional case. However, the relationship between network loadings and true population loadings in the dynamical case has yet to be fully investigated. The present study reveals an asymmetrical relationship between incremental changes in true population loadings and corresponding changes in network loadings, posing challenges to their interpretation. We propose adjustments to network loadings, which preserve the rank order of loadings while increasing their spread, enhancing the interpretability and discriminability of network structures.
Building the structure of personality from the bottom-up using Hierarchical Exploratory Graph Analysis
Andrew Samo, Alexander P Christensen, Francisco J Abad, Luis Eduardo Garrido, Marcos Jiménez, Eduardo Garcia-Garzon, Hudson Golino, Samuel T McAbee
Abstract
Understanding the structure of personality is fundamental for efforts to describe, predict, and explain it. Many recent calls have suggested that personality should be investigated from the bottom-up, as item-specific variance captures meaningful differences in personality that is often obscured by aggregation. Using a large, open-source 300-item IPIP-NEO dataset (N= 149,337), we employed a novel network psychometrics approach to capture personality from the bottom-up. The approach started with identifying sets of locally dependent items using Unique Variable Analysis to remove statistical artifacts that could hinder accurate dimension recovery. This analysis yielded 239 items that were then subjected to a novel, iterative adaptation of Hierarchical Exploratory Graph Analysis to identify the dimensions of personality level-by-level. This method identified 30 first-level dimensions, 6 second-level dimensions, and 2 third-level dimensions. These dimensions were compared against the theoretical assignments of the IPIP-NEO. Although there was considerable overlap, there were many items not assigned to their theoretical facets and several items not assigned within their theoretical trait domains. One strength of our approach was that items and dimensions were allowed to covary freely and these covariances were represented at each level of the trait hierarchy. At each level, we compare our results with the theoretical taxonomy of personality and discuss the promise of our novel approach to understand personality from the bottom-up.
Evaluating the structure of the Aesthetic Responsiveness Assessment (AReA) with Bootstrap Exploratory Graph Analysis
Paul J Silvia, Rebekah M Rodriguez-Boerwinkle, Kim N Awa, Darya L Zabelina, Alexander P Christensen
Some Terms
bootEGA
Bootstrap Exploratory Graph Analysis
GLASSO
graphical least absolute shrinkage and selection operator
LASSO
least absolute shrinkage and selection operator
GGM
Gaussian Graphical Model
EBIC
extended Bayesian information criterion
TMFG
triangulated maximally filtered graph
PA
parallel analysis
PAF
principal axis factoring
PCA
principal component analysis
ARI
Adjusted Rand Index
PC
percent correct
NMI
normalized mutual information
SSC
Simons Simplex Collection
BAPQ
Broad Autism Phenotype Questionnaire
,
, , ,
