Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses
The DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index.
Abstract
In this paper, we develop a classical approach to model selection. Using the Kullback-Leibler Information Criterion to measure the closeness of a model to the truth, we propose simple likelihood-ratio based statistics for testing the null hypothesis that the competing models are equally close to the true data generating process against the alternative hypothesis that one model is closer. The tests are directional and are derived successively for the cases where the competing models are non-nested, overlapping, or nested and whether both, one, or neither is misspecified. As a prerequisite, we fully characterize the asymptotic distribution of the likelihood ratio statistic under the most general conditions. We show that it is a weighted sum of chi-square distribution or a normal distribution depending on whether the distributions in the competing models closest to the truth are observationally identical. We also propose a test of this latter condition.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
In the citing paper
“A comparison of the two models using Vuong's distinguishability test showed that the eight factor model was a significantly better fit to the data than the single factor model (LR = 47,277.2, p < 0.001) (30).”
From Assessing the sustainability capacity of evidence-based programs in community and health settings (2022) · cites this paper as [30]
In the citing paper
“Nonetheless, as I and many others have argued [1,32-40], sequence similarity is strictly an empirical observation; homology, on the other hand, is a hypothesis intended to explain the similarity.”
From On universal common ancestry, sequence similarity, and phylogenetic structure: the sins of P-values and the virtues of Bayesian evidence (2011) · cites this paper as [40]
In the citing paper
“The Likelihood Ratio Test suggests that a 2-class solution fit better than a 1-class solution (p < .001), a 3-class solution fit better than a 2-class solution (p = .020), and a 4-class solution fit better than a 3-class solution (p = .025); however, a 5-class solution did not fit significantly better than a 4-class solution (p = .413).”
From Mindfulness and emotional outcomes: Identifying subgroups of college students using latent profile analysis (2015) · cites this paper as Vuong, 1989
In the citing paper
“Given the results of the Likelihood Ratio Test, we settled on the 4-class solution.”
From Mindfulness and emotional outcomes: Identifying subgroups of college students using latent profile analysis (2015) · cites this paper as Vuong, 1989
In the citing paper
“We find that we can reject the null hypothesis that the two models are equally close to the actual model in favor of zero-inflation, which is a superior model (p<0.01) [52].”
From Incorporating Anthropogenic Influences into Fire Probability Models: Effects of Human Activity and Climate Change on Fire Activity in California (2016) · cites this paper as [52]
Citations
Cites
A new look at the statistical model identification
Akaike, H. (1974) · IEEE Transactions on Automatic Control · doi:10.1109/tac.1974.1100705 · Cited by 51,879
Estimating the Dimension of a Model
Schwarz, G. (1978) · The Annals of Statistics · doi:10.1214/aos/1176344136 · Cited by 47,033
On Information and Sufficiency
Kullback, S., & Leibler, R. A. (1951) · The Annals of Mathematical Statistics · doi:10.1214/aoms/1177729694 · Cited by 20,193
The Large-Sample Distribution of the Likelihood Ratio for Testing Composite Hypotheses
Wilks, S. S. (1938) · The Annals of Mathematical Statistics · doi:10.1214/aoms/1177732360 · Cited by 3,493
Wald, A. (1943) · Transactions of the American Mathematical Society · doi:10.1090/s0002-9947-1943-0012401-3 · Cited by 2,511
Several Tests for Model Specification in the Presence of Alternative Hypotheses
Davidson, R., & MacKinnon, J. G. (1981) · Econometrica · doi:10.2307/1911522 · Cited by 1,747
Linear Statistical Inference and its Applications.
Sprent, P., & Rao, C. R. (1974) · Journal of the Royal Statistical Society. Series A (General) · doi:10.2307/2345151 · Cited by 1,747
Asymptotic Properties of Non-Linear Least Squares Estimators
Jennrich, R. I. (1969) · The Annals of Mathematical Statistics · doi:10.1214/aoms/1177697731 · Cited by 1,327
ON THE USE AND INTERPRETATION OF CERTAIN TEST CRITERIA FOR PURPOSES OF STATISTICAL INFERENCE
NEYMAN, J., & PEARSON, E. S. (1928) · Biometrika · doi:10.1093/biomet/20a.3-4.263 · Cited by 1,070
Further Results on Tests of Separate Families of Hypotheses
Cox, D. R. (1962) · Journal of the Royal Statistical Society Series B: Statistical Methodology · doi:10.1111/j.2517-6161.1962.tb00468.x · Cited by 1,026
Cited by
Items 201–216 of 216
Mendes, I., & Proença, I. (2011) · Environmental Management · doi:10.1007/s00267-011-9733-1 · Cited by 32
Faraone, S. V., & Glatt, S. J. (2010) · Journal of Attention Disorders · doi:10.1177/1087054709332472 · Cited by 30
A Longitudinal Examination of Enterprise Resource Planning System Post-Implementation Enhancements
Cao, J., Nicolaou, A. I., & Bhattacharya, S. (2013) · Journal of Information Systems · doi:10.2308/isys-50398 · Cited by 29
“Furthermore, untabulated Vuong (1989) tests show that industry competition and market share have significantly more explanatory power than other firm characteristics in the adoption decision.”
BackgroundSkripkauskaite, S., Fazel, M., & the OxWell Study Team (2022) · JMIR Pediatrics and Parenting · doi:10.2196/41480 · Cited by 29
Kim, S., Lee, J., & Boone, D. (2022) · Journal of Pain Research · doi:10.2147/jpr.s340382 · Cited by 26
Karlas, T., Petroff, D., Feisthammel, J., Beer, S., Blüher, M., Schütz, T., Lichtinghagen, R., Hoffmeister, A., & Wiegand, J. (2022) · Obesity Surgery · doi:10.1007/s11695-022-06150-5 · Cited by 26
Assessing the sustainability capacity of evidence-based programs in community and health settings
Bacon, C., Malone, S., Prewitt, K., Hackett, R., Hastings, M., Dexter, S., & Luke, D. A. (2022) · Frontiers in Health Services · doi:10.3389/frhs.2022.1004167 · Cited by 22
Vowles, K. E., Sowden, G., Hickman, J., & Ashworth, J. (2019) · Behaviour Research and Therapy · doi:10.1016/j.brat.2018.10.012 · Cited by 24
Secondhand smoke exposure in the workplace
Skeer, M., Cheng, D. M., Rigotti, N. A., & Siegel, M. (2005) · American Journal of Preventive Medicine · doi:10.1016/j.amepre.2005.01.007 · Cited by 23
Héroux, J., Moodie, E. E. M., Strumpf, E., Coyle, N., Tousignant, P., & Diop, M. (2014) · Statistics in Medicine · doi:10.1002/sim.6020 · Cited by 22
“Vuong’s likelihood ratio test for non-nested likelihoods [22] could also have been used, however its performance for weighted likelihoods in the semi-parametric setting of MSMs has not been investigated.”
MethodAgulnik, A., Malone, S., Puerto-Torres, M., Gonzalez-Ruiz, A., Vedaraju, Y., Wang, H., Graetz, D., Prewitt, K., Villegas, C., Cardenas-Aguierre, A., et al. (2021) · BMJ Open · doi:10.1136/bmjopen-2021-053116 · Cited by 20
Theobald, D. L. (2011) · Biology Direct · doi:10.1186/1745-6150-6-60 · Cited by 16
“The low data-toparameter ratio invalidates the AIC and non-nested LLR tests, as it strongly violates the approximations needed to justify the tests (the AIC and LLR are strictly only valid in the asymptotic limit of infinite data) [94,105].”
MethodMercier, F., Witczak, L. R., & Bales, K. L. (2020) · American Journal of Primatology · doi:10.1002/ajp.23141 · Cited by 15
“Because we only evaluated parameter significance in one model per outcome, posthoc corrections were not necessary (Vuong, 1989; West et al., 2012).”
MethodProfiles of Pain Acceptance and Values-Based Action in the Assessment and Treatment of Chronic Pain
Kruger, E., Ashworth, J., Sowden, G., Hickman, J., & Vowles, K. E. (2022) · The Journal of Pain · doi:10.1016/j.jpain.2022.06.005 · Cited by 5
Lee, J. Y., & Walton, D. M. (2021) · Canadian Journal of Pain · doi:10.1080/24740527.2020.1870102 · Cited by 5
Risk factors for injury among federal wildland firefighters in the United States
Britton, C. L. (2018) · doi:10.17077/etd.iog4m1n5 · Cited by 3
Metadata
- Authors
- Vuong, Q. H.
- Affiliations
- University of Southern California (US)
- Corresponding author
- Vuong, Quang H.
- Venue
- Econometrica
- Publisher
- JSTOR
- Published
- 1989-03
- Volume
- 57
- Issue
- 2
- Pages
- 307
- ISSN
- 0012-9682 (ISSN-L)
- OpenAlex
- W1830719945
- Cited by
- 5,929
- References
- 29
- Citations by year
- 2026: 178 · 2025: 197 · 2024: 185 · 2023: 213 · 2022: 264 · 2021: 331 · 2020: 289 · 2019: 325 · 2018: 345 · 2017: 287 · 2016: 296 · 2015: 343 · 2014: 357 · 2013: 328 · 2012: 283
- FWCI
- 15.00
- Type
- journal article
- Language
- en
- Open access
- closed
- Locations
- Econometrica (publishedVersion)
- CaltechAUTHORS (California Institute of Technology) (submittedVersion)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Inference
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Applications
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Bayesian Inference
- Keywords
- likelihood ratio test, model selection, Kullback-Leibler information criterion, model misspecification, asymptotic distribution, chi-square distribution, normal distribution, directional tests, nested models
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.
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
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.