Non-nested Models Research Articles

The effect of political and bureaucratic regime changes on Australia's real interest rate

We attempt to find out if changes in political and bureaucratic regimes lead to significant mean shifts in the Australian short-term real interest rate, using quarterly data for the period 1970:3 to 2019:1. Utilising a sophisticated structural break test, three distinct structural breaks are found in Australia's real interest rate. The empirical results show that a model estimated to consider the shifts in the political regime explains more of the variation in the real interest rate compared to a model that incorporates shifts in the bureaucratic regime. Results from several non-nested model selection tests also confirm this finding.

Modified harmonic mean method for spatial autoregressive models

In this paper, I suggest using the modified harmonic mean method for estimating marginal likelihood functions of cross-sectional spatial autoregressive models. In a Bayesian estimation setting, I show how this method can be used for popular cross-sectional spatial autoregressive models. In a simulation study, I investigate the finite sample performance of this estimator along with some other popular information criteria for the nested and non-nested model selection problems. The simulation results show that the modified harmonic mean estimator performs satisfactorily, and can be useful for the specification search exercises in spatial econometrics.

ФОРМАЛЬНЫЕ ТЕСТЫ ВЫБОРА СПЕЦИФИКАЦИИ МОДЕЛИ

The paper examines the construction of an econometric model of the dependence of the money supply aggregate M0 on nominal wages according to Russian Federation data. The problem of choosing a model specification is solved within the framework of formal tests, the algorithms of which are implemented in the R software environment. In practice, as a rule, the true model is unknown and as a result, a model is estimated that only approximately corresponds to the process that generates the data. To date, a number of approaches have been developed for testing the correctness of the choice of specification, both for nested and nonnested competing models. To test nested models, hypotheses are formulated in the form of restrictions on parameters, which are tested, for example, using a standard Ftest for comparing long (unrestricted) and short (restricted) regression models. Comparison of nonnested models is based on the formation of artificial embedding models – hybrid models that are tested using a nonnested Ftest or Jtest (Davidson, Mackinnon) [3, 9]. The Jtest uses only one constraint to make a reasonable choice of model, which leads to an increase in its power compared to a nonnested Ftest with a large number of additional regressors [2]. To select a functional form from a set of competing nonnested models, the BoxCox procedure is used, based on a formalized algorithm for selecting a linearizing transformation from a wide family of powerlaw transformations [1]. To compare linear and loglinear models, a simplified procedure proposed by Zarembka Paul [6] has found wide practical application. When constructing the econometric model under study using the listed procedures for selecting a functional form, the set of competing models included: a paired linear regression model; loglinear model, distributed lag model and autoregressive model.

Bayesian Solutions for the Factor Zoo: We Just Ran Two Quadrillion Models

ABSTRACTWe propose a novel framework for analyzing linear asset pricing models: simple, robust, and applicable to high‐dimensional problems. For a (potentially misspecified) stand‐alone model, it provides reliable price of risk estimates for both tradable and nontradable factors, and detects those weakly identified. For competing factors and (possibly nonnested) models, the method automatically selects the best specification—if a dominant one exists—or provides a Bayesian model averaging–stochastic discount factor (BMA‐SDF), if there is no clear winner. We analyze 2.25 quadrillion models generated by a large set of factors and find that the BMA‐SDF outperforms existing models in‐ and out‐of‐sample.

Optimal Model Averaging of Mixed-Data Kernel-Weighted Spline Regressions

Model averaging has a rich history dating from its use for combining forecasts from time-series models (Bates and Granger) and presents a compelling alternative to model selection methods. We propose a frequentist model averaging procedure defined over categorical regression splines (Ma, Racine, and Yang) that allows for mixed-data predictors, as well as nonnested and heteroscedastic candidate models. We demonstrate the asymptotic optimality of the proposed model averaging estimator, and develop a post-averaging inference theory for it. Theoretical underpinnings are provided, finite-sample performance is evaluated, and an empirical illustration reveals that the method is capable of outperforming a range of popular model selection criteria in applied settings. An R package is available for practitioners (Racine).

Herd level economic comparison between the shape of the lactation curve and 305 d milk production.

Herd milk production performance is generally evaluated using the herd's average 305-day milk production (HM305). Economic comparisons between herds are also often made using HM305. Comparing herds is thus based on summarized milk production, and not on the form of the lactation curves of the cows within the herd. Cow lactation curve characteristics can be aggregated on a calendar year basis to herd lactation curve characteristics (HLCC) (herd magnitude, herd time to peak yield and herd persistency). Thus far, no literature has evaluated whether the shape of the lactation curve (described by HLCC) is better able to explain the economic variation of herds than summarized milk production such as HM305 does. This study aims to determine whether HM305 or HLCC is better able to explain the variation in economic performance between herds. To do so, we evaluated 8 years of Dutch longitudinal data on milk production and the financial accounts of 1,664 herds. Cow lactation curve characteristics were calculated through lactation curve modeling and aggregated to HLCC on a calendar year basis for two parity groups (primiparous cows and multiparous cows). Using income over feed cost per cow (IOFC-cow) or per 100 kg milk (IOFC-milk) as the dependent variable separately, we developed four linear mixed models. Two models were used to analyse the association between herd economic performance and HLCC; the other two models were used to analyse the association between herd economic performance and HM305. A Cox test and J test were used to compare two non-nested models to investigate whether HM305 or HLCC better explain IOFC. The average IOFC-cow was €2,305 (SD = 408) per year, while the average IOFC-milk was €32.1 (SD = 4.6). Results showed that HLCC and HM305 explain the same amount of variance of IOFC-cow or IOFC-milk. IOFC-cow was associated with HM305 and HLCC (except herd time to peak yield for primiparous cows). Herd magnitude was most strongly associated with IOFC-cow, followed by herd persistency and herd time to peak yield of multiparous cows. IOFC-milk was not associated with HM305 or HLCC (except for a weak negative association with herd persistency for primiparous cows). IOFC-cow and IOFC-milk were driven most by time effects. In conclusion, HLCC and HM305 explain the same amount of variance in IOFC-cow or IOFC-milk. HLCC is more computationally expensive, while HM305 is more readily available.

Fast and Accurate Estimation of Non-Nested Binomial Hierarchical Models Using Variational Inference

Non-linear hierarchical models are commonly used in many disciplines. However, inference in the presence of non-nested effects and on large datasets is challenging and computationally burdensome. This paper provides two contributions to scalable and accurate inference. First, I derive a new mean-field variational algorithm for estimating binomial logistic hierarchical models with an arbitrary number of non-nested random effects. Second, I propose “marginally augmented variational Bayes” (MAVB) that further improves the initial approximation through a step of Bayesian post-processing. I prove that MAVB provides a guaranteed improvement in the approximation quality at low computational cost and induces dependencies that were assumed away by the initial factorization assumptions. I apply these techniques to a study of voter behavior using a high-dimensional application of the popular approach of multilevel regression and post-stratification (MRP). Existing estimation took hours whereas the algorithms proposed run in minutes. The posterior means are well-recovered even under strong factorization assumptions. Applying MAVB further improves the approximation by partially correcting the under-estimated variance. The proposed methodology is implemented in an open source software package.

Comparison of prognostic reliability of TNM staging systems version 7/8 and SARCULATOR in primary localized liposarcomas of the extremity or trunk wall.

e23529 Background: The oncological staging system should provide meaningful information regarding the patients’ prognosis. For soft tissue sarcomas, available staging systems include AJCC version 8, AJCC version 7, and nomogram-based approaches such as SARCULATOR (which additionally includes age and subtype). This work compares these systems using a group of extremity/trunk wall cases of liposarcoma (LPS). Methods: We have included all (N = 147) consecutive, newly diagnosed LPS patients treated in our institution between 2010 and 2021. We have excluded atypical lipomatous tumors. Patients’ and tumor characteristics were collected, and AJCC staging and SARCULATOR risk assessment were performed. Appropriate Cox proportional hazard models were fitted. The results were compared using the concordance index, Akaike information criterion (AIC), and partial likelihood ratio (PLR) test for non-nested Cox models. Results: In this group, 81 (55.1%) patients were male; the mean age in the group was 54 years (SD:16). Most tumors (83%) were localized on the lower extremity, with 8.8% on the upper extremity and 8.2% on the trunk. 66% of cases were myxoid LPS, 17.7% - dedifferentiated and 16.3% - pleomorphic. The median tumor size was 12 cm (IQR 9 – 18). Patients’ prognostic groups are summarized in Table. The median follow-up was 43 months (95%CI:38-51). The median overall survival (OS) was not reached. The 5-year survival rate was 68% (95%CI:62-75). The concordance for AJCC version 7, AJCC version 8, and SARCULATOR were respectively 0.625, 0.705, and 0.778. The AIC in the same order were: 311, 301, 274. The PLR test showed that AJCC version 8 fits better than AJCC version 7 (p for superiority = 0.036) while SARCULATOR fits better than AJCC version 8 (p for superiority = 0.028). Conclusions: New staging systems improve in better predicting patients outcome. In rare diseases such as sarcomas, it is crucial to have access to more personalized nomograms such as SARCULATOR.[Table: see text]

Modeling Time to Blindness of Glaucoma Patients: A Case Study at Jimma University Medical Center.

Glaucoma is a significant public health problem due to its substantial increase in the projected number of glaucoma cases. In Ethiopia, glaucoma accounts for 5.2% of irreversible blindness and is the fifth main cause of blindness. The main objective of this study was to modeling time to blindness of left and right eyes of glaucoma patients. An institution-based retrospective cohort study. This study was conducted among 315 glaucoma patients admitted to the Ophthalmology Department of Jimma University Medical Center (JUMC), Southwest Ethiopia, from January 1, 2016, to August 30, 2020. Kaplan-Meier survival analysis and semiparametric and parametric copula models were applied to identify factors that affect time to the blindness in glaucoma patients and the dependence between time to the blindness of the left and right eyes, respectively. An Akaike information criterion (AIC) was used to select the best non-nested model. In total, 211 (66.9%) out of 315 glaucoma patients were blind, whereas 104 (33.1%) patients were censored. The median time to the blindness of the left and right eyes was determined to be 12 months. The result suggested that the risk of the blindness in male patients was 1.005 (P = 0.01) times higher than that in female patients, and the risk of the blindness in patients who had early, moderate, and advanced glaucoma was estimated to be 0.582 (P = 0.002), 0.485 (P = 0.001) and 0.887 (P = 0.003) times less than that in the patients with absolute glaucoma, respectively. Age, place of residence, gender, type of medication, diabetes disease, stage of glaucoma, duration of treatment, intraocular pressure (IOP), and cup-disk ratio were significantly associated with and affected by the time to the blindness of left and right eyes in glaucoma patients. Awareness should be given to the community to reduce the burden of glaucoma.

Comparative validation of the GRACE 2.0 and EUROSCORE II for 6-month mortality risk estimation of patients with acute coronary syndrome undergoing urgent bypass graft surgery

Abstract Funding Acknowledgements Type of funding sources: Public grant(s) – National budget only. Main funding source(s): The research was supported by the Ministry of Innovation and Technology NRDI Office within the framework of the Artificial Intelligence National Laboratory Program. Background According to the current NSTE-ACS guidelines, the GRACE 2.0 score is used for predicting mortality risk of patients with acute coronary syndrome regardless of treatment modality. However, mortality risk of patients requiring acute bypass graft surgery is usually assessed by the EUROSCORE II model. Purpose We aimed to validate comparatively the predictive ability of the two models in NSTE-ACS patients undergoing urgent bypass surgery. Methods We analysed data of 99 complete cases of an unicentre registry who underwent acute CABG surgery between 2013 and 2021. Patients were followed-up by means of hospital records, follow-up visits, and records of the National Health Insurance Fund. Besides analysis of the bootstrapped c-statistics, we constructed univariate Cox regression models and compared model fit using Fine’s partial likelihood ratio test for non-nested models. Statistical analysis was performed using R. Results Of the ninety-nine cases, two patients were lost to follow-up at 180 days (2.02%). The mortality rate was 16/97 (16.49%) with a c-statistic of 0.82 (95% CI: 0.70 to 0.92) and 0.72 (95% CI: 0.57 to 0.86) for the GRACE 2.0 and EUROSCORE II models, respectively (p=0.11). Analysis of the fitted univariate Cox models revealed that the two models are highly ‘distinguishable’ (p=0.0004, variance test). Yet, with this small sample size, the difference in model fit did not reach statistical significance in the non-nested partial likelihood ratio test (p=0.08, that GRACE 2.0 fits better than EUROSCORE II). Conclusion The GRACE 2.0 score may have a good predictive power to estimate 180-day mortality risk of acute coronary syndrome patients undergoing acute bypass surgery. Our preliminary data may raise the possibility that it also may outperform the EUROSCORE II. Nevertheless, to test this hypothesis with adequate statistical power, larger, preferably multicentre data are needed.

Inequalities in utilization of institutional care among older people in Estonia

The purpose of this study is to shed light on various sources of inequalities in the utilization of institutional care among the Estonian elderly. We apply two-part hurdle and zero-inflated negative binomial models to analyze the association between various socioeconomic determinants and other factors on the one hand and the use of inpatient care and rehabilitation care services on the other. We discriminate between the two types of models using the new approach to comparing nonnested models – the HPC test. Applying the Lewbel estimator, we also explore the association between the utilization of inpatient care and rehabilitation. The evidence supports the existence of a pro-rich and pro-educated gradient in the needs-adjusted utilization of inpatient and rehabilitation care services among older people in Estonia. At the same time, we find no tangible urban/rural or ethnic barriers to the use of these services. The obtained results suggest the necessity of facilitating the accessibility of rehabilitation care services by mitigating both demand-side and supply-side barriers.

The Topp-Leone Gompertz-G family of distributions with applications

A new extension of the Topp-Leone distribution is developed to produce a new family of distributions, namely Topp-Leone Gompertz-G distribution. Some of its structural properties including expansion of density, distribution of order statistics, Renyi entropy, moments, quantile, and generating functions are derived. We considered four special cases of the new distribution and we conclude that the distribution can handle heavy tailed data and non-monotonic hazard rate functions. A simulation study was conducted to examine the performance of the special case of the proposed family of distributions. Two lifetime data examples are included in this article to illustrate the applicability of the proposed model in comparison with other competing non-nested models.

A corrected Clarke test for model selection and beyond

We introduce a large family of model selection tests based on the expectation of an arbitrary, possibly non-smooth, parametric criterion function of the data. The considered methodology is illustrated for several econometric problems, including linear and quantile regression. It covers the case of strictly locally non-nested models and some overlapping models. The asymptotic theory of the proposed test statistic is stated. A general exchangeable bootstrap scheme allows the evaluation of its limiting law as well as its asymptotic variance. Our framework includes the tests for non-nested model selection of Vuong (1989) and Clarke (2007) as particular cases. We show that the statistic of the latter test is not Binomial distributed as originally stated and we provide its corrected limiting law. In a simulation study, we empirically verify the distributional approximation of our test statistic in a finite sample and examine the empirical level and power of the corresponding model selection tests in various settings. Finally, an analysis of a financial dataset illustrates the proposed model selection procedure at work.

The odd power generalized Weibull-G power series class of distributions: properties and applications

Abstract We develop a new class of distributions, namely, the odd power generalized Weibull-G power series (OPGW-GPS) class of distributions. We present some special classes of the proposed distribution. Structural properties, have also been derived. We conducted a simulation study to evaluate the consistency of the maximum likelihood estimates. Moreover, two real data examples on selected data sets, to illustrate the usefulness of the new class of distributions. The proposed model outperforms several non-nested models on selected data sets.

Computation and application of generalized linear mixed model derivatives using lme4.

Maximum likelihood estimation of generalized linear mixed models (GLMMs) is difficult due to marginalization of the random effects. Derivative computations of a fitted GLMM's likelihood are also difficult, especially because the derivatives are not by-products of popular estimation algorithms. In this paper, we first describe theoretical results related to GLMM derivatives along with a quadrature method to efficiently compute the derivatives, focusing on fitted lme4 models with a single clustering variable. We describe how psychometric results related to item response models are helpful for obtaining the derivatives, as well as for verifying the derivatives' accuracies. We then provide a tutorial on the many possible uses of these derivatives, including robust standard errors, score tests of fixed effect parameters, and likelihood ratio tests of non-nested models. The derivative computation methods and applications described in the paper are all available in easily obtained R packages.

Unintended Consequences of Environmental Policies: the Case of Set-aside and Agricultural Intensification

The objective of this paper is to evaluate the indirect impacts of a set-aside policy on the environment through crop intensification. We estimate a structural, multi-output model on a panel of French farmers from 2006 to 2010, accounting for multivariate selection (corner solutions) on crops and land use. We use a parametric and a semi-nonparametric (based on sieve estimation) version of a quasi-maximum likelihood (QML) estimator, compare their goodness of fit and parameter efficiency and check for closeness with a test for non-nested models. We then perform a simulation experiment based on elasticity estimates for set-aside area and input demand, to evaluate intensive margin effects of the set-aside policy. Results show that a set-aside subsidy can provide farmers with incentives to intensify their production, leading to potential adverse environmental effects.

Confirmatory factor analysis of the Clinician-Administered PTSD Scale (CAPS-5) based on DSM-5 vs. ICD-11 criteria

ABSTRACT Introduction Many studies have investigated the latent structure of the DSM-5 criteria for posttraumatic stress disorder (PTSD). However, most research on this topic was based on self-report data. We aimed to investigate the latent structure of PTSD based on a clinical interview, the Clinician-Administered PTSD Scale (CAPS-5). Method A clinical sample of 345 participants took part in this multi-centre study. Participants were assessed with the CAPS-5 and the Posttraumatic Stress Disorder Checklist (PCL-5). We evaluated eight competing models of DSM-5 PTSD symptoms and three competing models of ICD-11 PTSD symptoms. Results The internal consistency of the CAPS-5 was replicated. In CFAs, the Anhedonia model emerged as the best fitting model within all tested DSM-5 models. However, when compared with the Anhedonia model, the non-nested ICD-11 model as a less complex three-factor solution showed better model fit indices. Discussion We discuss the findings in the context of earlier empirical findings as well as theoretical models of PTSD.

Continuous marketing in the social business platform based on identifying consumer boring products

In the social business platform, continuous marketing to consumers can fully explore the consumer purchasing potential. However, since consumers can be influenced by their social friends, their tastes often change, which resulting in the cold start problem of familiar users (CSPFU), and the traditional product recommendation methods are difficult to achieve satisfactory results because they focus on identifying the preferable new products instead of boring familiar products. Therefore, a consumer multi-stage compensation product evaluation model (CMCPEM) based on the multidimensional correlation of products and customers to identify the products that consumers may feel tired is proposed. Specifically, the multidimensional correlation indexes are firstly proposed to depict the preferences of the consumer for the target product to be identified, other consumers who have social contagion and structural equivalence relationships with the consumer and other consumers of homogeneous products. After the direct linear, non-linear and indirect fusion of these multidimensional correlation indexes, the compensation indexes (CIs) are proposed to comprehensively describe the first stage of product evaluation process of consumers. Then, J test in the non-nested model is used to screen out the non-nested CIs that consumers focus on. Finally, in the third stage, the final decision result is given by comprehensively considering CIs that consumers focus on and the indexes that represent consumers’ favorite. Experiment results on YELP data confirm the effectiveness of CMCPEM in successfully launching the continuous marketing campaign.

Type II Exponentiated Half-Logistic-Topp-Leone-G Power Series Class of Distributions with Applications

This paper aims to develop a new class of distributions, namely, type II exponentiated half-logistic Topp-Leone power series (TIIEHL-TL-GPS) class of distributions. Some important properties including moments, quantiles, moment generating function, entropy and maximum likelihood estimates are derived. A simulation is conducted study to evaluate the consistency of the maximum likelihood estimates. We also present three real data examples to illustrate the usefulness of the new class of distributions. Results shows that the proposed model performs better than nested and several non-nested models on selected data sets

Objective Bayesian variable selection in linear regression model

Variable selection in a regression model with k potential explanatory variables requires the choosing of a model among the possible submodels, which is a difficult task when the number of explanatory variables is moderately large. In this study, we propose the objective Bayesian variable selection procedures where the encompassing of the underlying nonnested linear models is crucial. Based on the encompassed models, objective priors for the multiple testing problem involved in the variable selection problem can be defined. The proposed approach provides a considerable reduction in the size of the compared models by restricting the posterior search for the right models, from to only k + 1, given k explanatory variables. Furthermore, the consistency of the proposed variable selection procedures was checked and their performance was examined using real examples and simulation analyzes by comparing the classical and Bayesian procedures of search in all possible submodels.