The DOI matches its Crossref record; 98 of the 98 references it cites were found in a registry or an index; 20 of the 93 sources its claims cite were checked for support.
Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations
Open accessAbstract
Many constructs in management studies, such as perceptions, personalities, attitudes, and behavioral intentions, are not directly observable. Typically, empirical studies measure such constructs using established scales with multiple indicators. When the scales are used in a different population, the items are translated into other languages or revised to adapt to other populations, it is essential for researchers to report the quality of measurement scales before using them to test hypotheses. Researchers commonly report the quality of these measurement scales based on Cronbach’s alpha and confirmatory factor analysis results. However, these results are usually inadequate and sometimes inappropriate. Moreover, researchers rarely consider sampling errors for these psychometric quality measures. In this best practice paper, we first critically review the most frequently-used approaches in empirical studies to evaluate the quality of measurement scales when using structural equation modeling. Next, we recommend best practices in assessing reliability, convergent and discriminant validity based on multiple criteria and taking sampling errors into consideration. Then, we illustrate with numerical examples the application of a specifically-developed R package, measureQ, that provides a one-stop solution for implementing the recommended best practices and a template for reporting the results. measureQ is easy to implement, even for those new to R. Our overall aim is to provide a best-practice reference for future authors, reviewers, and editors in reporting and reviewing the quality of measurement scales in empirical management studies.
Overview
CoChat Summary
TL;DR
This best-practice paper reviews how management researchers report reliability, convergent validity, and discriminant validity in SEM studies, argues that common practices (Cronbach's alpha, single-threshold rules of thumb) are inadequate and ignore sampling error, and recommends multi-criteria assessments using bootstrapped confidence intervals. It introduces the R package `measureQ` and illustrates it with five simulated examples plus a reporting template.
Key findings
- Cronbach's alpha is the most commonly reported reliability coefficient in SEM studies despite being non-SEM-based; its tau-equivalence assumption is "typically not justified," and it typically underestimates reliability when factor loadings are unequal. The authors recommend construct reliability (CR; McDonald's omega) instead, with reliability not significantly lower than 0.7 and ideally not significantly lower than 0.8.
- For second-order factors, omega total (ωT) represents reliability of the combined first-order factors, whereas omega hierarchical (ωh) is the more appropriate second-order reliability measure because it treats both residual and group-specific error as error.
- Convergent validity should require all three of Hair et al.'s (2009) conditions — CR ≥ 0.7, standardized loadings ≥ 0.5, AVE ≥ 0.5 — but tested with 90% bootstrapped CIs rather than as fixed cutoffs. In Example A, CRs ranged 0.8206 (IV) to 0.8944 (EB); IV's AVE was 0.4819 but not significantly below 0.5.
- Discriminant validity should be judged on four criteria: established convergent validity, no cross-loading, AVE greater than shared variance, and correlations assessed against 0.7, 0.8, and 0.85 using the lower limit of the 90% CI. The authors argue discriminant validity is a matter of degree, not a dichotomous decision.
- The paper identifies an error in Henseler et al.'s (2015) HTMT equation and provides a corrected formula (Eq. 10) using covariances rather than inter-item correlations; the corrected HTMT equals the correlation between simple-averaged composite scores disattenuated by Cronbach's alpha.
- In Example D (HR flexibility), the second-order AVE was 0.2634 and CFHRP's AVE was 0.4207 (significantly below 0.5), so convergent validity failed for the second-order HRF scale; the authors recommend testing hypotheses with only the four first-order factors that passed. In Example E, engagement–absorption's AVE (0.4937) was significantly lower than its squared correlation with engagement–dedication (0.8543² = 0.7298), raising discriminant validity concern.
Methods
- Best-practice/review paper (not an empirical study): critical review of frequently used approaches to reliability, convergent, and discriminant validity in covariance-based SEM, summarized in Table 1 (current practices and concerns) and Table 2 (recommended normative and operational definitions).
- Recommendations are implemented in a newly developed R package, `measureQ`, which uses bootstrapping (parametric bootstrapping for nested data), the MLR estimator, FIML for missing data, and 90% percentile CIs; default 1,000 bootstrapped samples.
- Five simulated examples illustrate the package: Example A (basic four-factor model, based on Yu et al. 2020, n = 354), B (single-item factor, based on Zahoor et al. 2022, n = 232), C (second-order factors, based on Lythreatis et al. 2022, n = 487), D (second-order HR flexibility, based on Way et al. 2015 Sample 3, n = 221), and E (second-order factor with nested data and two single-item factors, based on Wayne et al. 2017, n = 192 employees nested in 160 supervisors).
Limitations
- Stated: recommendations apply only to empirical studies using established measures in the multitrait-monomethod context, not scale development research in the multitrait-multimethod context; only to covariance-based SEM, not PLS-SEM; and only to reflective measures, not formative measures.
- Stated: because CIs are used, adequate sample size matters for power; following Anderson and Gerbing (1988) the authors recommend a minimum sample size of 150, and researchers should check for substantially wide CIs in smaller samples.
- Evident from the text: all five worked examples use simulated datasets rather than real data (the authors confirm the example data are "explicitly simulated for this study"), so the illustrations demonstrate the procedure rather than validate the criteria empirically; the paper also notes that results from two `measureQ` trials may differ slightly due to bootstrapping.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
13 Valid · 10 Not valid · 57 Not checked · 134 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“First, intending to estimate the disattenuated construct score correlation, Henseler et al.'s (2015, Eq. 6) HTMT equation is incorrect.”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
What the sources say
- Disputedcites Henseler et al. (2015) sourceExperimental — accuracy not yet measured
The paper does not say its Eq. 6 is incorrect. In the HTMT discussion (page unavailable), it describes HTMT as an estimate of the inter-construct correlation that parallels the disattenuated construct score correlation; the Appendix (page unavailable) says it demonstrates that HTMT is an estimator of that correlation. The paper thus presents the equation as valid rather than identifying the error alleged in the citing sentence. (LLM, conf=0.86)
Checked by a reasoning model — may be wrong.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“Ultimately, they did not recommend HTMT inference , given it failed to detect discriminant validity violations even when inter-construct correlations were as high as 0.95.”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
What the sources say
- Disputedcites Henseler et al. (2015) sourceExperimental — accuracy not yet measured
The Specificity results report that HTMT inference often treats constructs as distinct even when their correlation reaches 0.95, supporting the stated concern about its liberality. But the Key findings and recommendations explicitly recommend HTMT inference as an option in some settings, contrary to the claim that the authors did not recommend it. Page numbers are unavailable in our copy. (LLM, conf=0.91)
Checked by a reasoning model — may be wrong.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Hence, unless the construct scores are calculated by averaging the standardized item scores, Henseler et al.'s HTMT equation provides a biased estimate of the disattenuated correlation between two composites.”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
What the sources say
- Partly supportedcites Henseler et al. (2015) sourceExperimental — accuracy not yet measured
The HTMT discussion (section “b) a)”; page not provided) says HTMT estimates the correlation between constructs and parallels a disattenuated construct-score correlation. It also says HTMT does not require construct scores. The cited paper does not state that HTMT is biased for correlations between composites unless scores are calculated by averaging standardized items. (LLM, conf=0.85)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Second, both Henseler et al.'s (2015) HTMT equation and the corrected HTMT equation in Eq. ( 10) do not allow single-item factors because the denominator will be undefined if one or both constructs are single-item measures.”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
What the sources say
- Supportedcites Henseler et al. (2015) source
The Appendix derivation (page not provided) gives Henseler et al.’s HTMT equation terms containing Kᵢ(Kᵢ−1) and Kⱼ(Kⱼ−1). If either construct has one item, the corresponding term is undefined, supporting the limitation attributed to their equation. The citing paper’s corrected Eq. (10) is not attributed to Henseler et al. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“While Henseler et al.'s (2015) HTMT equation uses inter-item correlations instead of inter-item covariances, that becomes the disattenuated correlation between two composite scores formed by averaging the standardized item scores.”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
What the sources say
- Partly supportedcites Henseler et al. (2015) sourceExperimental — accuracy not yet measured
The empirical example and Appendix define HTMT using inter-item correlations, and the main discussion says it parallels a disattenuated construct-score correlation. The paper does not state that HTMT becomes the disattenuated correlation of two composites formed by averaging standardized item scores; it distinguishes HTMT from a method that requires calculating construct scores. (LLM, conf=0.85)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
96 more claims · 3 with their cited sources checked
Create a free account to see all 214 claims and every source verdict.
- Probably not a substantial claim. 70
- It may state two separate claims as one. 61
- The quote only partly states this claim. 4
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Values above 0.50 are generally considered acceptable cutoff points [63].”
From Psychometric validation and measurement invariance of the Health-Promoting Lifestyle Profile II in Honduran university students (2025) · cites this paper as [63]
Statistics recomputed
Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.
- Consistentχ²(3) = 662.09reported p < .01 · recomputed p < .001
“χ 2 (3) = 662.09,p < .01”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Compare the model fit of the unconstrained model versus the constrained model formed by combing the items of two factors into one factor (Anderson & Gerbing, 1988;Bagozzi & Phillips, 1982;Jöreskog, 1971) ➢ Fails to test discriminant validity for all pairs of constructs ➢ An overly sensitive test such that a correlation slightly lower than unity will erroneously show discriminant validity # The correlation between two constructs is less than 0.9, 0.85, 0.8, or 0.75 (Garson, 2002;Kenny, 2016) ➢ Ignores sampling errors ➢ When the correlation between two constructs is fixed to a constant, this fails to constrain the correlations between these two constructs and other constructs # The average variance extracted is greater than the shared variance (AVE-SV approach) (Fornell & Larcker, 1981) ➢ High AVEs tolerate high correlation between constructs ➢ Relies on a rule of thumb and ignores sampling errors # HTMT less than 0.85 (Henseler et al., 2015) ➢ Ignores measurement errors ➢ Assumes the factors are composites formed with standardized item scores ➢ Does not allow for single-item factors > Table 2
“# HTMT less than 0.85 (Henseler et al., 2015) ➢ Ignores measurement errors”
§ Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations > Heterotrait-monotrait (HTMT) ratio of correlations less than 0.85
“While Henseler et al.'s (2015) HTMT equation uses inter-item correlations instead of inter-item covariances, that becomes the disattenuated correlation between two composite scores formed by averaging the standardized item scores.”
Citations
Cites
Evaluating Structural Equation Models with Unobservable Variables and Measurement Error
Fornell, C., & Larcker, D. F. (1981) · Journal of Marketing Research · doi:10.2307/3151312 · Cited by 109,314
Evaluating structural equation models with unobservable variables and measurement error
Fornell, C., & Larcker, D. F. (1981) · Journal of Marketing Research · doi:10.1177/002224378101800104 · Cited by 109,190 · cited 14×
“The AVE should not be lower than 0.5 to demonstrate an acceptable level of convergent validity, meaning that the latent construct explains no less than 50% of the indicator variance (Fornell & Larcker, 1981, p. 46).”
InfluentialHu, L., & Bentler, P. M. (1999) · Structural Equation Modeling A Multidisciplinary Journal · doi:10.1080/10705519909540118 · Cited by 108,788 · cited 1×
“When the overall model fit indices indicate that the hypothesized measurement model fits the data adequately (refer to Hu & Bentler, 1999; Kline, 2016), this establishes the essential requirement for evaluating reliability, convergent and discriminant validity.”
Structural equation modeling in practice: A review and recommended two-step approach
Anderson, J. C., & Gerbing, D. W. (1988) · Psychological Bulletin · doi:10.1037/0033-2909.103.3.411 · Cited by 41,941 · cited 13×
“First proposed by Jöreskog (1971), with subsequent recommendations by Bagozzi and Phillips (1982, p. 476) and by Anderson and Gerbing (1988, p. 416), this is the most widely-used approach to evaluate discriminant validity.”
InfluentialAn Introduction to the Bootstrap
Efron, B., & Tibshirani, R. J. (1994) · doi:10.1201/9780429246593 · Cited by 40,500
Principles and Practice of Structural Equation Modeling
Tabri, N., & Elliott, C. M. (2012) · Canadian Graduate Journal of Sociology and Criminology · doi:10.15353/cgjsc-rcessc.v1i1.25 · Cited by 38,891
A new criterion for assessing discriminant validity in variance-based structural equation modeling
Henseler, J., Ringle, C. M., & Sarstedt, M. (2015) · Journal of the Academy of Marketing Science · doi:10.1007/s11747-014-0403-8 · Cited by 38,459 · cited 8×
“Second, both Henseler et al.’s (2015) HTMT equation and the corrected HTMT equation in Eq.”
Influentiallavaan : An R Package for Structural Equation Modeling
Rosseel, Y. (2012) · Journal of Statistical Software · doi:10.18637/jss.v048.i02 · Cited by 27,955 · cited 1×
“In addition to measureQ, conducting CFA requires the R package lavaan (Rosseel, 2012), so both measureQ and lavaan should be installed once on the computer.”
Alternative Ways of Assessing Model Fit
Browne, M. W., & Cudeck, R. (1992) · Sociological Methods & Research · doi:10.1177/0049124192021002005 · Cited by 27,200 · cited 1×
“…(Appelbaum et al., 2018) refer researchers to Kline (2016, Chapter 18), who suggested researchers report chi-square with its degrees of freedom and p -value; RMSEA (Steiger & Lind, 1980) and its 90% confidence interval (Browne & Cudeck, 1992), CFI (Bentler, 1990), and SRMR (Bentler, 1995).”
Comparative fit indexes in structural models.
Bentler, P. M. (1990) · Psychological Bulletin · doi:10.1037/0033-2909.107.2.238 · Cited by 25,266 · cited 7×
“The results for Example C provided in the supplementary file include the fit indices of the hypothesized second-order factor model, which show that 2 with 367 degrees of freedom = 399.40, RMSEA = 0.013, CFI = 0.997, and SRMR = 0.013, indicating that our measurement model fits the data well.”
Influential
Research works printed but not linked: 2
- Bentler, P. M. (1995). EQS 5 [Computer Program]. Multivariate Software Inc.
- Nunnally, J. C. (1978). Psychometric theory (2nd ed.). McGraw-Hill.
Cited by
Kassa, B. Y., & Worku, E. K. (2025) · Journal of Open Innovation Technology Market and Complexity · doi:10.1016/j.joitmc.2025.100474 · Cited by 263
Kalkbrenner, M. T. (2024) · Measurement and Evaluation in Counseling and Development · doi:10.1080/07481756.2023.2283637 · Cited by 174
Enhancing public service delivery efficiency: Exploring the impact of AI
Kulal, A., Rahiman, H. U., Suvarna, H., Abhishek, N., & Dinesh, S. (2024) · Journal of Open Innovation Technology Market and Complexity · doi:10.1016/j.joitmc.2024.100329 · Cited by 203
The relationship between inert thinking and ChatGPT dependence: An I-PACE model perspective
Ye, J., Zhang, M., Nong, W., Wang, L., & Yang, X. (2024) · Education and Information Technologies · doi:10.1007/s10639-024-12966-8 · Cited by 87
Self-citationEntrepreneurial education and intention: fear of failure, self-efficacy and gender
Duong, C. D., & Xuan, V. N. (2023) · Journal of Small Business and Enterprise Development · doi:10.1108/jsbed-02-2023-0057 · Cited by 85
Metadata
- Authors
- Cheung, G. W., Cooper-Thomas, H. D., Lau, R. S., & Wang, L. C.
- Affiliations
- University of Auckland (NZ)
- Auckland University of Technology (NZ)
- Hang Seng University of Hong Kong (HK)
- Corresponding author
- Cheung, Gordon W.
- Venue
- Asia Pacific Journal of Management
- Published
- 2023-01-30
- Volume
- 41
- Issue
- 2
- Pages
- 745-783
- ISSN
- 0217-4561 (ISSN-L)
- 1572-9958
- OpenAlex
- W4318457133
- Semantic Scholar
- 69075fbef6c034d3d32f95fc72c514e2b4a14244
- Cited by
- 2,262
- References
- 94
- Citations by year
- 2026: 912 · 2025: 895 · 2024: 374 · 2023: 80
- FWCI
- 415.33
- Type
- journal article
- Language
- en
- Update notices
- correction (2023-03-25)
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0 (tdm)
- creativecommons.org/licenses/by/4.0 (vor)
- Locations
- Asia Pacific Journal of Management (publishedVersion, cc-by)
- Tuwhera (Auckland University of Technology) (submittedVersion, cc-by)
- RePEc: Research Papers in Economics (submittedVersion)
- Topics
- Social Sciences › Decision Sciences › Management Science and Operations Research › Psychometric Methodologies and Testing
- Social Sciences › Decision Sciences › Statistics, Probability and Uncertainty › Reliability and Agreement in Measurement
- Social Sciences › Business, Management and Accounting › Strategy and Management › Business Strategies and Management Research
- Keywords
- reliability statistics, convergent validity, discriminant validity, structural equation modeling, sampling error, Cronbach's alpha, confirmatory factor analysis, psychometrics
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations. https://app.cochat.ai/papers/10.1007/s10490-023-09871-y
- The paper
- Cheung, G. W., Cooper-Thomas, H. D., Lau, R. S., & Wang, L. C. (2024). Reporting reliability, convergent and discriminant validity with structural equation modeling: A review and best-practice recommendations. Asia Pacific Journal of Management. https://doi.org/10.1007/s10490-023-09871-y
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.