The problem, known in psychometrics as method variance or the wording effect, arises because scales are frequently built with a deliberate balance of positively and negatively keyed items. The idea sounds sensible: reversing the direction of some questions should discourage automatic agreement and force respondents to read carefully. But the strategy backfires in a measurable way. Items that share the same wording direction tend to correlate with one another for reasons that have nothing to do with the underlying construct. When researchers run a factor analysis on such data, the statistical machinery often obligingly splits the scale into two clusters, one of positively worded items and one of negatively worded items, rather than the single dimension the scale was designed to capture.

This artificial splitting, sometimes called spurious bidimensionality, has real consequences. A researcher who concludes that a questionnaire measures two distinct traits may publish misleading findings, build theories on phantom subfactors, or compute reliability estimates that are inflated or deflated for the wrong reasons. The grit scale, a wildly popular measure of perseverance and passion for long-term goals, has been a particular battleground for this debate. Since its development by Angela Duckworth and colleagues, researchers have argued endlessly about whether grit truly has two facets, perseverance of effort and consistency of interest, or whether the apparent two-factor structure is simply an artifact of the negatively worded items embedded in the short version of the scale.

Faraci and Nasonte tackled this question with random intercept item factor analysis, or RIIFA, a modeling approach originally introduced by Albert Maydeu-Olivares and Donna Coffman in 2006. The core insight of RIIFA is elegant: it adds a random intercept factor to the standard factor model, a latent variable on which every item in the scale loads regardless of its wording direction. This intercept factor absorbs the shared response tendency that cuts across all items, including acquiescence, the general willingness to agree with statements, and other response styles. Once that pervasive method variance is siphoned off into its own latent dimension, the remaining substantive factor can be interpreted as the true trait, purified of the contamination introduced by how the questions happen to be phrased.

To evaluate whether RIIFA delivers on this promise, the researchers conducted two studies using two independent samples from the United Kingdom. The first study analyzed data from 977 participants, and the second from 496. Their test case was the Short Grit Scale, known as Grit-S, an eight-item instrument that mixes positively and negatively worded questions. The choice was strategic: the Grit-S is short, widely used, and famously prone to the wording-effect problem, making it an ideal proving ground for a method intended to restore structural validity to mixed-worded scales.

The first study focused on dimensionality assessment, the task of determining how many latent factors actually underlie a set of items. The researchers compared traditional factor retention techniques, including parallel analysis, a classic method that compares observed eigenvalues against those from random data, with RIIFA-based counterparts that incorporate the random intercept factor. They also employed exploratory graph analysis, a newer network psychometrics approach that treats items as nodes in a network and uses community detection algorithms, borrowed originally from network science methods like the fast unfolding of communities, to identify clusters of tightly connected items. The results were striking. Traditional retention methods consistently overestimated the number of factors, suggesting the Grit-S contained two dimensions when the theoretical expectation was one. In contrast, the RIIFA-based techniques produced unidimensional solutions, and bootstrap analyses, which resample the data thousands of times to gauge stability, showed that these solutions were considerably more stable than those obtained without controlling for the wording effect.

The second study shifted from exploration to confirmation, using confirmatory factor analysis to formally test competing models of the scale’s structure. The researchers fit models with and without a random intercept factor and compared them on both fit and parsimony, the principle that simpler explanations should be preferred unless complexity earns its keep. The model incorporating the random intercept factor achieved the best balance between the two. Its fit statistics were excellent by conventional standards: a root mean square error of approximation of .048, with a confidence interval spanning .022 to .072, a comparative fit index of .984, a Tucker-Lewis index of .974, and a standardized root mean square residual of .027. Values close to .95 or above on the fit indices and below about .06 or .08 on the error measures are typically considered indicative of a well-fitting model, and this model cleared every bar comfortably.

Reliability told a similarly encouraging story. The RIIFA model yielded a hierarchical omega of .84, a coefficient derived from bifactor-style measurement models that estimates how well the total scale score reflects a single dominant common factor after accounting for the variance absorbed by subsidiary dimensions. In practical terms, this means that once the method variance was reallocated to the random intercept factor, the substantive grit factor explained enough common variance to support meaningful interpretation of total scores. The researchers interpret this as evidence that RIIFA does not merely hide the problem; it actively redistributes explained variance between the substantive factor and the method factor, mitigating the artificial bidimensionality that plagues conventional analyses and enhancing the interpretability of the latent structure.

The implications reach well beyond the grit scale. Mixed-worded instruments are ubiquitous across psychology, appearing in measures of self-esteem, core self-evaluations, need for cognition, perceived stress, learning burnout, and countless clinical and organizational questionnaires. For each of these, the same dilemma recurs: researchers must decide whether an emerging second factor represents a genuine substantive distinction or a methodological artifact. Historically, that judgment has been made with tools, such as standard parallel analysis and conventional confirmatory factor analysis, that have no built-in mechanism for separating content from phrasing. The findings of this study suggest that RIIFA offers a principled alternative, and the authors explicitly recommend its application in cases where wording effects threaten the validity of psychometric measurement.

The study also connects to a broader movement in quantitative psychology toward modeling response styles and careless responding rather than simply hoping they wash out in large samples. Recent work has documented how even a few inconsistent respondents can confound the structure of personality survey data, how acquiescence distorts exploratory factor analysis, and how attention checks and response-style detection methods carry their own complications. Faraci and Nasonte’s contribution fits squarely into this research program, demonstrating on real data that a model explicitly designed to partition substantive and method variance can outperform traditional approaches. Importantly, the authors have made their data openly available through the Open Science Framework, and all of the R and Mplus code used in the analyses is accessible as well, lowering the barrier for other researchers to adopt the technique. For a field that depends so heavily on self-report questionnaires, a validated method for ensuring that scores reflect traits rather than question phrasing is not a technical nicety. It is a foundation for the credibility of the measurements themselves, and this study provides compelling empirical grounds for adding random intercept item factor analysis to the standard psychometric toolkit.

Subject of Research: Using random intercept item factor analysis to separate substantive variance from wording-effect method variance in mixed-worded psychological scales

Article Title: Disentangling substantive and method variance in mixed-worded scales: An empirical application of the random intercept item factor analysis (RIIFA)

Article References: Faraci, P., & Nasonte, G. (2026). Disentangling substantive and method variance in mixed-worded scales: An empirical application of the random intercept item factor analysis (RIIFA). Behavior Research Methods, 58(11), Article 299. https://doi.org/10.3758/s13428-026-03163-1

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03163-1

Keywords: psychometrics, random intercept item factor analysis, wording effects, method variance, Short Grit Scale, factor analysis, exploratory graph analysis, parallel analysis, dimensionality, structural validity, confirmatory factor analysis, survey methodology

For decades, clinicians and researchers have grouped the heterogeneous experiences of schizophrenia into broad domains. Positive symptoms such as hallucinations and delusions were separated from negative symptoms like blunted affect and social withdrawal, while disorganized speech and behavior formed a third cluster. These divisions, formalized in instruments such as the Positive and Negative Syndrome Scale, shaped drug development trials, genetic association studies, and the diagnostic criteria in the DSM and ICD. Yet the assumption that these categories reflect distinct underlying biological mechanisms has never been firmly established, and critics have argued that the dimensions are artifacts of the measurement instruments rather than discoveries about the illness itself.

Graph analysis, the approach at the heart of the new study, treats symptoms as nodes in a network and statistical associations between them as edges. Instead of asking whether a set of preassigned items load onto a particular factor, exploratory graph analysis applies algorithms derived from network science to detect communities of tightly interconnected symptoms. Methods such as the walktrap algorithm combined with regularization techniques can identify clusters that emerge purely from the pattern of relationships in the data. Researchers can then compare these data-driven communities against the conventional dimensions to see whether the traditional structure holds up under scrutiny.

The appeal of this methodology lies in its ability to sidestep some of the assumptions baked into classical factor analysis. Exploratory graph analysis has been shown in simulation studies to recover the correct number of dimensions more reliably than older techniques, particularly when samples are moderate in size or when the underlying factors are correlated. In psychometrics more broadly, the technique has spread rapidly, finding applications in depression, personality research, and quality-of-life measurement. Its arrival in schizophrenia research signals a broader shift toward network approaches that view mental disorders as systems of interacting elements rather than reflections of single latent causes.

Applying these tools to symptom ratings from people with schizophrenia, the researchers examined whether positive, negative, and disorganized symptoms genuinely form separable communities, or whether alternative configurations better describe the clinical reality. The stakes of this question are considerable. If symptom dimensions overlap more than assumed, clinical trials that measure only one domain may miss treatment effects that ripple across the network. If, on the other hand, the dimensions are robust, they remain valid targets for precision psychiatry approaches that aim to match patients to treatments based on symptom profiles rather than categorical diagnoses.

The study also speaks to a persistent puzzle in schizophrenia genetics. Genome-wide association studies have identified hundreds of genetic variants that contribute to risk, but connecting those variants to specific symptom dimensions has proven difficult, with studies of symptom genetics often yielding inconsistent results. Part of the inconsistency may stem from measurement: if the dimensions themselves are unstable across cohorts, instruments, and statistical methods, then genetic analyses built on top of them inherit that instability. Establishing whether the classical structure is reproducible under modern, assumption-light methods is therefore a prerequisite for meaningful biological discovery.

Network perspectives bring additional conceptual benefits. They make it possible to identify bridge symptoms, items that connect otherwise separate communities and may act as pathways through which dysfunction spreads. In depression research, for example, bridge symptoms such as sleep disturbance have been proposed as links between anxiety and mood clusters. In schizophrenia, identifying which symptoms bridge positive and negative domains could point to intervention targets with the broadest downstream impact, and could help explain why some patients deteriorate along multiple dimensions simultaneously while others remain relatively circumscribed in their difficulties.

The methodological rigor demanded by graph analysis is also part of the story. Exploratory graph analysis relies on estimating a regularized partial correlation network, typically through the graphical least absolute shrinkage and selection operator, which sets small spurious associations to zero and leaves a sparse network whose community structure can be extracted. Stability checks, such as bootstrapped estimates of edge weights and centrality indices, are essential to ensure that the detected communities are not statistical mirages. Cross-validation across independent samples provides a further safeguard, and the field has increasingly insisted on such replication before accepting any new dimensional structure as credible.

What makes this study timely is the convergence of several trends in psychiatric science. The National Institute of Mental Health’s Research Domain Criteria initiative has pressed researchers to move beyond diagnostic categories toward dimensions grounded in behavior and biology. Meanwhile, large-scale datasets and improved computational tools have made it feasible to test the architecture of psychopathology with unprecedented statistical power. Revisiting the symptom dimensions of schizophrenia with contemporary network methods is a natural next step in this program, and the findings carry implications that extend from the clinic to the genetics laboratory.

For clinicians, the message is that the familiar dimensional map of schizophrenia remains a useful, but not infallible, guide. For researchers, the study demonstrates that the tools of network science can interrogate long-standing psychiatric constructs in ways that classical psychometrics could not, potentially revealing where the traditional categories deserve preservation and where they require revision. As the field moves toward biologically informed classification, studies of this kind serve as a bridge, testing whether the clinical vocabulary accumulated over a century of observation can withstand the scrutiny of modern data science, and helping to ensure that future research into the causes and treatments of schizophrenia rests on foundations that can bear the weight.

Subject of Research: Symptom dimensions in schizophrenia analyzed with exploratory graph analysis

Article Title: Revisiting symptom dimensions in schizophrenia with exploratory graph analysis

Article References: Illing, S., & Leucht, S. (2026). Revisiting symptom dimensions in schizophrenia with exploratory graph analysis. Schizophrenia, 12(1), Article 72. https://doi.org/10.1038/s41537-026-00799-y

Image Credits: AI Generated

DOI: 10.1038/s41537-026-00799-y

Keywords: schizophrenia, symptom dimensions, exploratory graph analysis, network analysis, psychometrics, positive symptoms, negative symptoms, disorganization, precision psychiatry, psychiatric classification, statistical methods, mental health

The research, conducted by Domenico Cangemi and Livia De Giovanni of LUISS Università di Roma together with Pierpaolo D’Urso and Vincenzina Vitale of Sapienza Università di Roma, applies an enhanced form of Exploratory Graph Analysis, or EGA, to the Varieties of Democracy dataset, the most ambitious expert-based measurement project in political science. V-Dem covers more than 600 indicators annually from 1789 to the present, drawing on evaluations from a pool of more than 4,200 country specialists who assess concepts that no official statistic can capture, such as academic freedom, judicial independence and the deliberative quality of public reasoning. For each country-year observation, the project typically collects assessments from five experts and feeds them into a Bayesian measurement model that estimates the reliability of each coder, corrects for systematic differences in how experts interpret response scales, and produces interval-level latent estimates together with credible regions that express measurement uncertainty.

From this vast apparatus the researchers selected 37 interval-valued indicators covering 179 countries in 2024, indicators that V-Dem’s theory organizes into five high-level principles of democracy: electoral, liberal, participatory, deliberative and egalitarian. These principles are operationalized through indices such as the Electoral Democracy Index, which combines clean elections with freedom of expression, association and universal suffrage, the Liberal Component Index, which captures judicial and legislative constraints on executive power, the Participatory Component Index, which measures civil society engagement and direct democracy, the Deliberative Component Index, which gauges whether decisions emerge from public reasoning oriented toward the common good, and the Egalitarian Component Index, which tracks equal access to political power and resources across social groups.

The core question was simple to state and hard to answer: does the empirically observed dependence among these 37 indicators reproduce the theoretically defined five-dimension structure, or does the data tell a different story? To find out, the team turned to EGA, a network psychometrics method in which variables appear as nodes and their conditional associations as edges, so that latent dimensions emerge as communities of densely connected nodes. The standard pipeline estimates a Gaussian Graphical Model using the graphical LASSO, a regularization technique that shrinks small partial correlations to zero, and then applies the Walktrap community detection algorithm, which finds clusters through random walks that preferentially traverse strong edges. The trouble is that this machinery quietly assumes multivariate normality, and political indicators are anything but normal. Measures of democracy remain substantially skewed or heavy-tailed across countries, and under such conditions the Pearson correlation matrix, the workhorse input of the standard method, can distort the estimated network and with it the recovered dimensions.

The study’s first methodological innovation addresses exactly this weakness. The authors replace the Pearson matrix with a nonparanormal estimator, a semiparametric device that treats each observed variable as a monotone transformation of a latent Gaussian variable, thereby preserving a Gaussian copula while allowing the marginals to take almost any shape. Under this model, the latent Gaussian correlation can be recovered directly from Spearman rank correlations through a trigonometric transformation, and this identity survives any strictly increasing rescaling of the data. In simulation experiments with three true latent dimensions and deliberately skewed, heavy-tailed marginals, the contrast was striking. At sample sizes of 50 and 100, the Pearson-based routine repeatedly failed to recover the correct number of dimensions, over-extracting four or more communities in 46 and 38 of 100 replications respectively, while the nonparanormal approach recovered the true three-dimensional structure nearly perfectly. Even at 500 observations the Pearson method still misclassified a variable, whereas the rank-based method did not.

Robustness to non-normality, however, solves only half the problem. A single EGA solution delivers a crisp partition of variables into communities, but both the estimated network and the resulting community assignments can wobble under small perturbations of the data, a concern that has shadowed network psychometrics since studies showed that psychopathology symptom networks can have limited replicability. The second innovation therefore wraps the analysis in a bootstrap. The team generated 500 synthetic datasets from a multivariate normal distribution built on the estimated nonparanormal correlation matrix, ran EGA on each, and then summarized the ensemble of 500 partitions through a co-association matrix, whose entries record the proportion of replications in which each pair of indicators landed in the same community. This consensus matrix is invariant to arbitrary relabeling of communities across runs, preserving the full ensemble information rather than reducing it to per-item stability scores.

Instead of forcing the consensus structure back into a crisp partition, the researchers applied Non-Euclidean Fuzzy Relational Clustering, an algorithm that works directly on the dissimilarity matrix derived from the co-association counts and produces fuzzy memberships, allowing each indicator to belong partially to several latent dimensions. Indicators with one dominant membership can be read as stable core components of a dimension, while those with distributed memberships flag genuine ambiguity or conceptual overlap. The authors are careful to stress that fuzziness is not necessarily a sign of model failure; it can reveal systematic cross-dimensional association that a binary assignment would simply erase.

When the full pipeline ran on the 2024 V-Dem data, the nonparanormal version identified five latent dimensions, exactly matching the theoretical classification, while the Pearson-based routine produced six and scored lower against the theory-based reference partition on the Adjusted Rand Index. The fuzzy pipeline built on the nonparanormal consensus achieved the highest fuzzy agreement of all tested variants. Most indicators behaved as theory predicted: all nine freedom of expression indicators clustered together, all five deliberative component indicators stayed intact, and most judicial constraint indicators formed a coherent group. But two findings stood out. First, freedom of religion, which V-Dem classifies under equality before the law, attached itself empirically to the freedom of expression cluster, consistent with the idea that countries suppressing religious practice tend also to suppress media and civil liberties. Second, and more substantively, the algorithm split V-Dem’s Equality before the Law and Individual Liberty index, which theory treats as a single construct, into two sharply distinct dimensions: one gathering institutional constraints on state behavior, including access to justice, freedom from torture, freedom from political killings, impartial administration and transparent law enforcement, and the other gathering individual freedoms with a pronounced gender component, including property rights, freedom of domestic movement and freedom from forced labor for men and women, alongside women’s civic participation.

That split carries a provocative implication: the rule-of-law securities a state provides and the personal freedoms its citizens actually experience do not necessarily move together. A regime may protect institutional legality while individual and gender-specific liberties lag behind, and the conventional index aggregates these divergent profiles into a single number, conflating regimes that are empirically quite different. Four indicators also emerged as fuzzy, including freedom of foreign movement, whose nearly uniform membership across all dimensions reflects how exit restrictions intertwine civil liberty, judicial transparency and political control at once. The authors propose practical consequences: index users could weight ambiguous indicators by their membership values, treat near-uniform ones as bridging indicators to be monitored separately, and compare regimes at the level of empirically recovered dimensions rather than relying solely on theoretical aggregation. The recovered structure survived sensitivity checks, holding when the community detection algorithm was swapped for Louvain or Leiden, when the fuzziness parameter varied, and largely when an alternative fuzzy clustering method was used. The team frames the rule-of-law split as an empirical suggestion worthy of dedicated investigation rather than an established property of governance, but the message for the growing industry of democracy monitoring is clear: behind every headline number lies a dimensional structure, and that structure, examined with the right statistical tools, can tell stories the headline number conceals.

Subject of Research: Robust network psychometric methods for recovering latent dimensions in cross-national democracy measurements

Article Title: Recovering Latent Dimensions in Cross-National Democracy Measurements Through Robust Exploratory Graph Analysis

Article References: Cangemi, D., D’Urso, P., De Giovanni, L., & Vitale, V. (2026). Recovering Latent Dimensions in Cross-National Democracy Measurements Through Robust Exploratory Graph Analysis. Social Indicators Research, 184(2), Article 41. https://doi.org/10.1007/s11205-026-03929-2

Image Credits: AI Generated

DOI: 10.1007/s11205-026-03929-2

Keywords: Exploratory Graph Analysis, V-Dem, democracy indices, nonparanormal model, network psychometrics, fuzzy clustering, Gaussian graphical models, rule of law, Spearman correlation, bootstrap stability, latent dimensions, Social Indicators Research