Search

AbstractA model-based approach is developed for clustering categorical data with no natural ordering. The proposed method exploits the Hamming distance to define a family of probability mass functions to model the data. The elements of this family are then considered as kernels of a finite mixture model with an unknown number of components. Conjugate Bayesian inference has been derived for the parameters of the Hamming distribution model. The mixture is framed in a Bayesian nonparametric setting, and a transdimensional blocked Gibbs sampler is developed to provide full Bayesian inference on the number of clusters, their structure, and the group-specific parameters, facilitating the computation with respect to customary reversible jump algorithms. The proposed model encompasses a parsimonious latent class model as a special case when the number of components is fixed. Model performances are assessed via a simulation study and reference datasets, showing improvements in clustering recovery over existing approaches. [ABSTRACT FROM AUTHOR]

This paper extends the literature on structural estimation of social preferences to account for the desire to adhere to social norms and hide one's true intentions via moral wiggle room. We conduct an experiment to test whether accounting for normatively appropriate behavior allows us to distinguish between preference types who care about outcomes versus adhering to social norms and whether the introduction of moral wiggle room undermines the stability of social preference estimates. We find that social preference estimates are remarkably robust to the inclusion of moral wiggle room. However, the representative agent is strongly motivated by norms and failing to account for this motive in our model causes us to overestimate how much agents care about helping those who are worse off. Using finite mixture models to endogenously identify latent preference types, we replicate previous work finding that the majority of subjects can be classified as strong or moderate altruists when normative concerns are not considered. Accounting for the normative appropriateness of decisions when categorizing participants, however, reveals different motives across types: strong altruists are only marginally concerned with norms while the moderate altruists are highly sensitive to them and, once norms are taken into account, don't care at all about the outcomes of others. Our results thus recast the prior findings in a new light. Rather than the two most common types being strong altruists and moderate altruists, we find that they are better described as strong altruists and norm followers. [ABSTRACT FROM AUTHOR]

We consider the Bayesian estimation of the parameters of a finite mixture model from independent order statistics arising from imperfect ranked set sampling designs. As a cost‐effective method, ranked set sampling enables us to incorporate easily attainable characteristics, as ranking information, into data collection and Bayesian estimation. To handle the special structure of the ranked set samples, we develop a Bayesian estimation approach exploiting the Expectation‐Maximization (EM) algorithm in estimating the ranking parameters and Metropolis within Gibbs Sampling to estimate the parameters of the underlying mixture model. Our findings show that the proposed RSS‐based Bayesian estimation method outperforms the commonly used Bayesian counterpart using simple random sampling. The developed method is finally applied to estimate the bone disorder status of women aged 50 and older. [ABSTRACT FROM AUTHOR]

The purpose of this research is to determine whether PISA 2018 mathematical literacy test items show a differential item functioning across countries. For this purpose, only the items in booklet number three were examined using the MIMIC method with Latent Class Analysis (LCA) approach. PISA 2018 tests are mostly developed in English. Therefore, in DIF analyses, the reference group is the UK, while the focal groups consist of the other countries examined in the research (Türkiye, Finland, Japan, and the USA). According to the results, of the 23 test items, statistically significant DIF was observed in eight items in the UK-Türkiye sample, in seven items in the UK-Finland sample, in eleven items in the UK-Japan sample, and in three items in the UK-USA sample. It is seen that the effect and size of DIF in non-homogeneous groups differ between groups and these effects can be examined in more detail with the LCA method. [ABSTRACT FROM AUTHOR]

The chemical characterization of materials recovered from archaeological sites has been used to relate artifacts to their respective sources of raw material. For this, conventional methods of classical statistics are commonly used, such as bivariate graphs, cluster analysis and linear data transformations. Those who have applied these methods claim to have a high degree of confidence in the correct assignment of materials to their respective sources. However, if empirical data deviate from theoretical assumptions, classical statistical techniques can produce incorrect assignments. In this paper, we propose a semi-supervised learning procedure using clustering and classification based on models of mixtures of multivariate asymmetric exponential power distributions. The objective is to trace the origin and provenance of obsidian archaeological materials from the archaeological zone of Izapa, Chiapas, significantly optimizing the result of the correct allocation of unlabeled data by employing only limited supervision in the form of labeled instances. The effectiveness of the proposed method was evaluated by a controlled experiment using data from a set of geological samples from obsidian deposits. Subsequently, the method was applied to analyze a set of obsidian artifacts collected in excavations within Izapa dated for the Classic period. In both cases, well-defined classifications were obtained that, in the case of Izapa, allowed to determine the consumption of materials from Guatemalan sources, as well as some relationship with sources from western and central Mexico. [ABSTRACT FROM AUTHOR]

The field of forensic statistics offers a unique hierarchical data structure in which a population is composed of several subpopulations of sources and a sample is collected from each source. This subpopulation structure creates an additional layer of complexity. Hence, the data has a hierarchical structure in addition to the existence of underlying subpopulations. Finite mixtures are known for modeling heterogeneity; however, previous parameter estimation procedures assume that the data is generated through a simple random sampling process. We propose using a semi‐supervised mixture modeling approach to model the subpopulation structure which leverages the fact that we know the collection of samples came from the same source, yet an unknown subpopulation. A simulation study and a real data analysis based on famous glass datasets and a keystroke dynamic typing data set show that the proposed approach performs better than other approaches that have been used previously in practice. [ABSTRACT FROM AUTHOR]

In archaeometry, the correct assignment of archaeological artefacts with their original geological sources continues to be an indispensable tool for determining exploitation patterns, trade routes, tool production and the movement of goods in ancient times. Statistically, Mahalanobis distance has been an integral part of the artefact provenance methodology. However, this method presents disadvantages when some of the theoretical assumptions are not fulfilled. In this way and with a focus on the veracity of the correct assignment of archaeological samples to their geological sources, we present a technique that employs a semi‐supervised classification based on finite mixture models. This procedure uses labelled and unlabelled data to assign samples to groups with the advantage that it does not depend on the sample size of the groups as required by Mahalanobis distance. This procedure was first applied in a controlled experiment where the exact origin of the samples is known in order to verify the accuracy of the method; afterwards, it was applied to artefacts recovered from the archaeological site of Xalasco, Tlaxcala (Mexico), from both surface and excavation collections and analysed with a portable X‐ray fluorescence spectrometer. The resulting groups defined that the ancient dwellers of Xalasco exploited six supply sources located in Central Mexico. [ABSTRACT FROM AUTHOR]

Investigations into household structure in low- and middle-income countries (LMICs) provide important insight into how families manage domestic life in response to resource allocation and caregiving needs during periods of rapid sociopolitical and health-related challenges. Recent evidence on household structure in many LMICs contrasts with long-standing viewpoints of worldwide convergence to a Western nuclearized household model. Here, we adopt a household-centered theoretical and methodological framework to investigate longitudinal patterns and dynamics of household structure in a rural South African setting during a period of high AIDS-related mortality and socioeconomic change. Data come from the Agincourt Health and Socio-Demographic Surveillance System (2003-2015). Using latent transition models, we derived six distinct household types by examining conditional interdependency between household heads' characteristics, members' age composition, and migration status. More than half of households were characterized by their complex and multigenerational profiles, with considerable within-typology variation in household size and dependency structure. Transition analyses showed stability of household types under female headship, while higher proportions of nuclearized household types dissolved over time. Household dissolution was closely linked to prior mortality experiences--particularly, following death of a male head. Our findings highlight the need to better conceptualize and contextualize household changes across populations and over time. [ABSTRACT FROM AUTHOR]

It is common to observe significant heterogeneity in clustered data across scientific fields. Cluster-wise conditional distributions are widely used to explore variations and relationships within and among clusters. This paper aims to capture such heterogeneity by employing cluster-wise finite mixture models. To address the heterogeneity among clusters, we introduce latent group structure and incorporate heterogeneous mixing proportions across different groups, accommodating the diverse characteristics observed in the data. The specific number of groups and their membership are unknown. To identify the latent group structure, we employ concave penalty functions to the pairwise differences of the preliminary consistent estimators for the mixing proportions. This approach enables the automatic division of clusters into finite subgroups. Theoretical results demonstrate that as the number of clusters and cluster sizes tend to infinity, the true latent group structure can be recovered with probability close to one, and the post-classification estimators exhibit oracle efficiency. We support our proposed approach’s performance and applicability through extensive simulations and analysis of basic consumption expenditure among urban households in China. [ABSTRACT FROM AUTHOR]

Gaussian mixture models are widely employed in serological data analysis to discern between seropositive and seronegative individuals. However, serological populations often exhibit significant skewness, making symmetric distributions like Normal or Student-t distributions unreliable. In this study, we propose finite mixture models based on Skew-Normal and Skew-t distributions for serological data analysis. Although these distributions are well established in the literature, their application to serological data needs further exploration, with emphasis on the determination of the threshold that distinguishes seronegative from seropositive populations. Our previous work proposed three methods to estimate the cutoff point when the true serological status is unknown. This paper aims to compare the three cutoff techniques in terms of their reliability to estimate the true threshold value. To attain this goal, we conducted a Monte Carlo simulation study. The proposed cutoff points were also applied to an antibody dataset against four SARS-CoV-2 virus antigens where the true serological status is known. For this real dataset, we also compared the performance of our estimated cutoff points with the ROC curve method, commonly used in situations where the true serological status is known. [ABSTRACT FROM AUTHOR]

Old-growth forests provide many ecosystem services and benefits. However, they are becoming increasingly rare and thus are an urgent priority for conservation. Accurately mapping old-growth forests is a critical step in this process. Here, we used LiDAR, an improved individual tree crown delineation algorithm for broadleaved forests, Gaussian mixture modelling, and a rule-based classification key to map the extent and location of old-growth forests across a topographically and ecologically complex landscape of 337,548 ha in southeastern Australia. We found that variation in old growth extent was largely driven by the old growth definition, which is a human construct, rather than by uncertainty in the technical aspect of the work. Current regulations define a stand as old growth if it was recruited prior to 1900 (i.e., >120 years old) and is undisturbed (i.e., <10% regrowth canopy cover and no visible disturbance traces). Only 2.7% (95% confidence intervals ranging from 1.4 to 4.9%) of the forests in the study landscape met these criteria. However, this definition is overly restrictive as it leaves many multi-aged stands with ecologically mature elements (e.g., one or more legacy trees amid regrowth) unprotected. Removing the regrowth filter, an indicator of past disturbances, increased the proportion of old-growth forests from 2.7% to 15% of the landscape. Our analyses also revealed that 60% of giant trees (>250 cm in diameter at breast height) were located within 50 m of cool temperate rainforests and cool temperate mixed forests (i.e., streamlines). We discuss the implication of our findings for the conservation and management of high-conservation-value forests in the region. [ABSTRACT FROM AUTHOR]

In this paper, we formulate a hierarchical Bayesian version of the Mixture of Unigrams model for text clustering and approach its posterior inference through variational inference. We compute the explicit expression of the variational objective function for our hierarchical model under a mean-field approximation. We then derive the update equations of a suitable algorithm based on coordinate ascent to find local maxima of the variational target, and estimate the model parameters through the optimized variational hyperparameters. The advantages of variational algorithms over traditional Markov Chain Monte Carlo methods based on iterative posterior sampling are also discussed in detail. [ABSTRACT FROM AUTHOR]

Identifying the strategies that are played is critical to understanding behavior in repeated games. This process is difficult because only choices (not strategies) are observable. Recently, a debate has emerged regarding whether subjects play mixed strategies in the indefinitely repeated prisoner's dilemma. We use an experimental approach to elicit mixed strategies from human subjects, thereby providing direct empirical evidence. We find that a majority of subjects use mixed strategies. However, the data also suggest subjects' strategies are becoming less mixed over time, and move toward three focal pure strategies: Tit For Tat, Grim Trigger, and Always Defect. We use the elicited strategies to provide an empirically‐relevant foundation for analyzing commonly used mixture model estimation procedures. [ABSTRACT FROM AUTHOR]

Gaussian mixture models, which assume a Normal distribution for each component, are popular in antibody (or serological) data analysis to help determining antibody-positive and antibody-negative individuals. In this work, we advocate using finite mixture models based on Skew-Normal and Skew-t distributions for serological data analysis. These flexible mixing distributions have the advantage of describing right and left asymmetry often observed in the distributions of known antibody-negative and antibody-positive individuals, respectively. We illustrate the application of these alternative mixture models in a data set on the role of human herpesviruses in the Myalgic Encephalomyelitis/Chronic Fatigue Syndrome.

Decisions in management and finance rely on information that often includes win-lose feedback (e.g., gains and losses, success and failure). Simple reinforcement then suggests to blindly repeat choices if they led to success in the past and change them otherwise, which might conflict with Bayesian updating of beliefs. We use finite mixture models and hidden Markov models, adapted from machine learning, to uncover behavioral heterogeneity in the reliance on difference behavioral rules across and within individuals in a belief-updating experiment. Most decision makers rely both on Bayesian updating and reinforcement. Paradoxically, an increase in incentives increases the reliance on reinforcement because the win-lose cues become more salient. This paper was accepted by Gustavo Manso, finance. Funding: C. Alós-Ferrer gratefully acknowledges financial support from the Deutsche Forschungsgemeinschaft under [Grant AL1169/4], part of the research unit "Psychoeconomics" (FOR 1882). Supplemental Material: The data files and online appendices are available at https://doi.org/10.1287/mnsc.2022.4584. [ABSTRACT FROM AUTHOR]

We introduce a bivariate bidimensional mixed‐effects regression model, motivated by the analysis of CO2$$ {\mathrm{CO}}_2 $$ emission levels and growth on OECD countries from 1990 to 2018. The model is able to capture heterogeneity across countries and allows for a full association structure among outcomes, assuming a discrete distribution for the random terms with a possibly different number of support points in each univariate profile. We test the behavior of the proposed approach via a simulation study, considering several factors such as the number of observed units, times, and levels of heterogeneity in the data. Empirically, we define an extended version of the STIRPAT model where all model parameters, and not only the mean, vary according to a regression model. Our empirical findings provide evidence of heterogeneous behaviors across countries and suggest the need of a flexible approach to properly reflect the heterogeneity in both the emission levels and the growth processes. [ABSTRACT FROM AUTHOR]

Empirical demand functions, such as those from choice-based conjoint analyses, are critical to many aspects of marketing. Approaches have been developed to ensure that research subjects provide honest and thoughtful responses. However, to reduce costs, researchers increasingly collect data online, under conditions that compromise the value of the information provided. Objective measures related to how the study is completed, such as latency (how quickly answers are given), can only be tied to other objective measures (such as the consistency of the answers), but ultimately their relationship to the subject's utility function is questionable. To address this problem, the authors introduce a mixture modeling framework that clusters subjects based on variances. The proposed model naturally groups subjects based on their internal consistency. The authors argue that a higher level of internal consistency (i.e., lower variance) reflects more engaged consumers who have sufficient experience with the product category and choice task. "Gremlins," in contrast, behave such that the noise in their responses overwhelms any signal, leading to a lack of predictive power. This approach provides an automated way to determine which respondents are relevant. The authors discuss the conceptual and modeling framework and illustrate the method using both simulated and commercial data. [ABSTRACT FROM AUTHOR]

The finite mixtures approach identifies homogeneous groups within the sample. The data are aggregated into classes sharing similar patterns without any prior knowledge or assumption on the clustering. These clusters are characterized by group-specific regression coefficients to account for between groups heterogeneity. Two different approaches have been independently defined in the literature to compute this estimator not only at the conditional mean but also in the tails. One approach allows the grouping to change according to the selected location. The other defines the clusters once and for all at the conditional mean, and then moves the estimation to the tails, focusing on cluster specific estimates and allowing between groups comparison. Here we compare the behavior of both approaches, and in addition we consider a closely related estimator based on expectiles, together with few others more robust, quantile-based estimators. A case study on students' performance concludes the analysis. [ABSTRACT FROM AUTHOR]

Jon Helgeland, Doris Tove Kristoffersen, Katrine Damgaard Skyrud Division for Health Services, Norwegian Institute of Public Health, Oslo, NorwayCorrespondence: Jon Helgeland, Norwegian Institute of Public Health, PO Box 222 Skøyen, Oslo, 0213, Norway, Tel +47 464 00 443, Email Jon.Helgeland@fhi.noObjective: Health registries are important data sources for epidemiology, quality monitoring, and improvement. Acute myocardial infarction (AMI) is a common, serious condition. Little is known about variation in the positive predictive value (PPV) of a coded AMI diagnosis and its association with hospital quality indicators. The present study aimed to investigate the relationship between PPV and registry-based 30-day mortality after AMI admission and between-hospital variation in PPV.Study Design and Setting: An electronic record review was performed in a nationwide sample of Norwegian hospitals. Clinical signs and cardiac troponin measurements were abstracted and analyzed using a mixture model for likelihood ratios and parametric bootstrapping.Results: The overall PPV was estimated to be 97%. We found no statistically significant association between hospital PPV and the classification of hospitals into low, intermediate, and high registry-based 30-day mortality. There was significant variation between hospitals, with a PPV range of 91– 100%.Conclusion: We found no evidence that variation in PPV of AMI diagnosis can explain variation between hospitals in registry-based 30-day mortality after admission. However, PPV varied significantly between hospitals. We were able to use a very efficient statistical approach to the analysis and handling of various sources of uncertainty.Keywords: health registries, quality indicators, finite mixture models, case fatality, cardiac troponins

Abstract Background Increasing attention is being given to assessing treatment effect heterogeneity among individuals belonging to qualitatively different latent subgroups. Inference routinely proceeds by first partitioning the individuals into subgroups, then estimating the subgroup-specific average treatment effects. However, because the subgroups are only latently associated with the observed variables, the actual individual subgroup memberships are rarely known with certainty in practice and thus have to be imputed. Ignoring the uncertainty in the imputed memberships precludes misclassification errors, potentially leading to biased results and incorrect conclusions. Methods We propose a strategy for assessing the sensitivity of inference to classification uncertainty when using such classify-analyze approaches for subgroup effect analyses. We exploit each individual’s typically nonzero predictive or posterior subgroup membership probabilities to gauge the stability of the resultant subgroup-specific average causal effects estimates over different, carefully selected subsets of the individuals. Because the membership probabilities are subject to sampling variability, we propose Monte Carlo confidence intervals that explicitly acknowledge the imprecision in the estimated subgroup memberships via perturbations using a parametric bootstrap. The proposal is widely applicable and avoids stringent causal or structural assumptions that existing bias-adjustment or bias-correction methods rely on. Results Using two different publicly available real-world datasets, we illustrate how the proposed strategy supplements existing latent subgroup effect analyses to shed light on the potential impact of classification uncertainty on inference. First, individuals are partitioned into latent subgroups based on their medical and health history. Then within each fixed latent subgroup, the average treatment effect is assessed using an augmented inverse propensity score weighted estimator. Finally, utilizing the proposed sensitivity analysis reveals different subgroup-specific effects that are mostly insensitive to potential misclassification. Conclusions Our proposed sensitivity analysis is straightforward to implement, provides both graphical and numerical summaries, and readily permits assessing the sensitivity of any machine learning-based causal effect estimator to classification uncertainty. We recommend making such sensitivity analyses more routine in latent subgroup effect analyses.

This paper aims to identify typologies of Latin American cities based on socioeconomic urban environment patterns. We used census data from 371 urban agglomerations in 11 countries included in the SALURBAL project to identify socioeconomic typologies of cities in Latin America. Exploratory factor analysis was used to select a set of variables, and finite mixture modelling (FMM) was applied to identify clusters to define the typology of cities. Despite the heterogeneities among the Latin American cities, we also found similarities. By exploring intersections and contrasts among these clusters, it was possible to define five socioeconomic regional typology patterns. The main features of each one are low-education cities in Northeast Brazil; low-unemployment cities in Peru and Panama; high-education cities in Argentina, Chile, Colombia, Costa Rica, Nicaragua and Mexico; high female labor participation, with high primary education in Argentina and low primary education in Brazil; and low female labor participation and low education in Brazil, Colombia, El Salvador, Guatemala, and Mexico. Identifying clusters of cities with similar features underscores understanding of the urban social and economic development dynamics and assists in studying how urban features affect health, the environment, and sustainability. [ABSTRACT FROM AUTHOR]

Finite mixture (FM) models have found key applications in many fields. Recently, some discussions have been made on comparing finite mixture models. In this paper, we discuss stochastic comparison of two FM models with respect to usual stochastic order when the mixture components have a general family of distributions. This problem has been studied when there is heterogeneity in one parameter (i.e., the distributional parameter), as well as when there is heterogeneity in two parameters (i.e., the distributional parameter and the mixing proportions). The sufficient conditions considered are based on p-larger order and reciprocally majorization order. Several examples have been provided to illustrate the established results. [ABSTRACT FROM AUTHOR]

Researchers use finite mixture models to analyze linked survey and administrative data on labor earnings, while also accounting for various types of measurement error in each data source. Different combinations of error-ridden and error-free observations characterize latent classes. Latent class probabilities depend on the probabilities of the different types of error. We introduce a suite of commands to fit finite mixture models to linked survey-administrative data: there is a general model and seven simpler variants. We also provide postestimation commands for assessment of reliability, marginal effects, data simulation, and prediction of hybrid variables that combine information from both data sources about the outcome of interest. Our commands can also be used to study measurement errors in other variables besides labor earnings. [ABSTRACT FROM AUTHOR]

There has been a substantial body of research on mixtures-of-regressions models that has developed over the past 20 years. While much of the recent literature has focused on flexible mixtures-of-regressions models, there is still considerable utility for imposing structure on the mixture components through fully parametric models. One feature of the data that is scantly addressed in mixtures of regressions is the presence of measurement error in the predictors. The limited existing research on this topic concerns the case where classical measurement error is added to the classic mixtures-of-linear-regressions model. In this paper, we consider the setting of mixtures of polynomial regressions where the predictors are subject to classical measurement error. Moreover, each component is allowed to have a different degree for the polynomial structure. We utilize a generalized expectation-maximization algorithm for performing maximum likelihood estimation. For estimating standard errors, we extend a semiparametric bootstrap routine that has been employed for mixtures of linear regressions without measurement error in the predictors. Numeric work, for practical reasons identified, is limited to estimating two-component models. We consider a likelihood ratio test for determining if there is a higher-degree polynomial term in one of the components. Model selection criteria are also highlighted as a way for determining an appropriate model. A simulation study and an application to the classic nitric oxide emissions data are provided. [ABSTRACT FROM AUTHOR]

This article presents a latent class model for multilevel data to identify latent subgroups and estimate heterogeneous treatment effects. Unlike sequential approaches that partition data first and then estimate average treatment effects (ATEs) within classes, we employ a Bayesian procedure to jointly estimate mixing probability, selection, and outcome models so that misclassification does not obstruct estimation of treatment effects. Simulation demonstrates that the proposed method finds the correct number of latent classes, estimates class-specific treatment effects well, and provides proper posterior standard deviations and credible intervals of ATEs. We apply this method to Trends in International Mathematics and Science Study data to investigate the effects of private science lessons on achievement scores and then find two latent classes, one with zero ATE and the other with positive ATE. [ABSTRACT FROM AUTHOR]

Gaussian mixture models are widely employed in serological data analysis to discern between seropositive and seronegative individuals. However, serological populations often exhibit significant skewness, making symmetric distributions like Normal or Student-t distributions unreliable. In this study, we propose finite mixture models based on Skew-Normal and Skew-t distributions for serological data analysis. Although these distributions are well established in the literature, their application to serological data needs further exploration, with emphasis on the determination of the threshold that distinguishes seronegative from seropositive populations. Our previous work proposed three methods to estimate the cutoff point when the true serological status is unknown. This paper aims to compare the three cutoff techniques in terms of their reliability to estimate the true threshold value. To attain this goal, we conducted a Monte Carlo simulation study. The proposed cutoff points were also applied to an antibody dataset against four SARS-CoV-2 virus antigens where the true serological status is known. For this real dataset, we also compared the performance of our estimated cutoff points with the ROC curve method, commonly used in situations where the true serological status is known.

Old-growth forests provide many ecosystem services and benefits. However, they are becoming increasingly rare and thus are an urgent priority for conservation. Accurately mapping old-growth forests is a critical step in this process. Here, we used LiDAR, an improved individual tree crown delineation algorithm for broadleaved forests, Gaussian mixture modelling, and a rule-based classification key to map the extent and location of old-growth forests across a topographically and ecologically complex landscape of 337,548 ha in southeastern Australia. We found that variation in old growth extent was largely driven by the old growth definition, which is a human construct, rather than by uncertainty in the technical aspect of the work. Current regulations define a stand as old growth if it was recruited prior to 1900 (i.e., >120 years old) and is undisturbed (i.e., 250 cm in diameter at breast height) were located within 50 m of cool temperate rainforests and cool temperate mixed forests (i.e., streamlines). We discuss the implication of our findings for the conservation and management of high-conservation-value forests in the region.

Economists and policy makers are interested in producers' responses to policies in order to achieve some national or sectoral objectives, e.g., growth, employment, food security. The way producers respond to policy depends on their production function. If producers do not have homogenous production function, policy responses will be heterogeneous. We use the underlying functional relationship to derive homogenous groupings. The paper employs finite regression mixture models to specify and estimate farm groups with regard to pre-specified functional relationship. The proposed approach is illustrated with regard to the aggregate production function of Kosovo agriculture, characterised by high prevalence of small farmers. The results point out to two farm clusters. The first one extracts more output from labour and intermediate consumption. The second one makes a better use of land. Perhaps, surprisingly, both clusters appear quite similar in terms of their stock of production inputs. Cluster 1 however appears to be more specialised. We can conclude that in Kosovo agriculture appearances and size are not primary determinants of productivity. [ABSTRACT FROM AUTHOR]

We develop and apply new statistical models for linked survey and administrative data on employment earnings, incorporating 4 types of measurement error. In addition, we allow error distributions to differ with individual characteristics, which improves model fit and allows us to investigate substantive hypotheses about factors associated with error bias and variance. Contributing the first UK evidence to a field dominated by findings about the USA, we show that measurement errors are pervasive, but the 4 types are quite different in nature. We also document substantial heterogeneity in each of the error distributions. [ABSTRACT FROM AUTHOR]

Partially recorded data are frequently encountered in many applications and usually clustered by first removing incomplete cases or features with missing values, or by imputing missing values, followed by application of a clustering algorithm to the resulting altered dataset. Here, we develop clustering methodology through a modelbased approach using the marginal density for the observed values, assuming a finite mixture model of multivariate t distributions. We compare our approximate algorithm to the corresponding full expectation–maximization (EM) approach that considers missing values in the incomplete dataset and makes a missing at random (MAR) assumption, as well as case deletion and imputation methods. Since only the observed values are utilized, our approach is computationally more efficient than imputation or full EM. Simulation studies demonstrate that our approach has favourable recovery of the true cluster partition compared to case deletion and imputation under various missingness mechanisms and is at least competitive with the full EM approach, even when MAR assumptions are violated. Our methodology is demonstrated on a problem of clustering gamma-ray bursts and is implemented at https://github.com/emilygoren/MixtClust. [ABSTRACT FROM AUTHOR]

Sustainability of agriculture is difficult to measure and assess because it is a multidimensional concept that involves economic, social and environmental aspects and is subjected to temporal evolution and geographical differences. Existing studies assessing agricultural sustainability in the European Union (EU) are affected by several shortcomings that limit their relevance for policy makers. Specifically, most of them focus on farm level or cover a small set of countries, and the few exceptions covering a broad set of countries consider only a subset of the sustainable dimensions or rely on cross-sectional data. In this paper, we consider yearly data on 12 indicators (5 for the economic, 3 for the social and 4 for the environmental dimension) measured on 26 EU countries in the period 2004–2018 (15 years), and apply group-based multivariate trajectory modeling to identify groups of countries with common trends of sustainable objectives. An expectation-maximization algorithm is proposed to perform maximum likelihood estimation from incomplete data without relying on an explicit imputation procedure. Our results highlight three groups of countries with distinguished strong and weak sustainable objectives. Strong objectives common to all the three groups include improvement of productivity, increase of personal income in rural areas, reduction of poverty in rural areas, increase of production of renewable energy, rise of organic farming and reduction of nitrogen balance. Instead, enhancement of manager turnover and reduction of greenhouse gas emissions are weak objectives common to all the three groups of countries. Our findings represent a valuable resource to formulate new schemes for the attribution of subsidies within the Common Agricultural Policy (CAP). [ABSTRACT FROM AUTHOR]

Finite mixture models are useful for modeling lifetimes distribution of random samples arising from heterogeneous populations. In this work, we establish some stochastic comparison results for two finite mixtures models with components having generalized Lehmann distribution. [ABSTRACT FROM AUTHOR]

Linear mixed-effects models are commonly used when multiple correlated measurements are made for each unit of interest. Some inherent features of these data can make the analysis challenging, such as when the series of responses are repeatedly collected for each subject at irregular intervals over time or when the data are subject to some upper and/or lower detection limits of the experimental equipment. Moreover, if units are suspected of forming distinct clusters over time, i.e., heterogeneity, then the class of finite mixtures of linear mixed-effects models is required. This paper considers the problem of clustering heterogeneous longitudinal data in a mixture framework and proposes a finite mixture of multivariate normal linear mixed-effects model. This model allows us to accommodate more complex features of longitudinal data, such as measurement at irregular intervals over time and censored data. Furthermore, we consider a damped exponential correlation structure for the random error to deal with serial correlation among the within-subject errors. An efficient expectation-maximization algorithm is employed to compute the maximum likelihood estimation of the parameters. The algorithm has closed-form expressions at the E-step that rely on formulas for the mean and variance of the multivariate truncated normal distributions. Furthermore, a general information-based method to approximate the asymptotic covariance matrix is also presented. Results obtained from the analysis of both simulated and real HIV/AIDS datasets are reported to demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

This paper proposes strict periodic stationarity and periodic ergodicity conditions for a finite mixture integer-valued GARCH model with S-periodic time-varying parameters that depend on the state of an independent and periodically distributed regime sequence. In this model, the past conditional mean values depend on the past of the regime variable in the same order, so the model is characterized by path-regime dependence. We also propose sufficient conditions for periodic stationarity when the conditional means are nonlinear of past observations. The results are applied to various discrete conditional distributions.

The clinical-high-risk for psychosis (CHR-P) syndrome is heterogeneous in terms of clinical presentation and outcomes. Identifying more homogenous subtypes of the syndrome may help clarify its etiology and improve the prediction of psychotic illness. This study applied latent class cluster analysis (LCCA) to symptom ratings from the North American Prodrome Longitudinal Studies 1 and 2 (NAPLS 1 and 2). These analyses produced evidence for three to five subgroups within the CHR-P syndrome. Differences in negative and disorganized symptoms distinguished among the subgroups. Subgroup membership was found to predict conversion to psychosis. The authors contrast the methods employed within this study with previous attempts to identify more homogenous subgroups of CHR-P individuals and discuss how these results could be tested in future samples of CHR-P individuals.

Finite mixture model-based clustering is a popular method to perform unsupervised learning, however the classic approach for parameter estimation, the expectation-maximization algorithm, is quite susceptible to converging to, at times extremely poor, local maxima. Recently, swarm-based algorithms have gained popularity in a number of optimization scenarios. One such swarm-based optimization procedure, the artificial bee colony, incorporates both local searching and multiple solution generating to explore the parameter space. Herein, we propose hybridizing these algorithms for the task of fitting a multivariate Gaussian mixture model. This algorithm's performance is contrasted with both the traditional approach and another hybrid algorithm on simulated and real data. [ABSTRACT FROM AUTHOR]

Background: Increasing attention is being given to assessing treatment effect heterogeneity among individuals belonging to qualitatively different latent subgroups. Inference routinely proceeds by first partitioning the individuals into subgroups, then estimating the subgroup-specific average treatment effects. However, because the subgroups are only latently associated with the observed variables, the actual individual subgroup memberships are rarely known with certainty in practice and thus have to be imputed. Ignoring the uncertainty in the imputed memberships precludes misclassification errors, potentially leading to biased results and incorrect conclusions.Methods: We propose a strategy for assessing the sensitivity of inference to classification uncertainty when using such classify-analyze approaches for subgroup effect analyses. We exploit each individual's typically nonzero predictive or posterior subgroup membership probabilities to gauge the stability of the resultant subgroup-specific average causal effects estimates over different, carefully selected subsets of the individuals. Because the membership probabilities are subject to sampling variability, we propose Monte Carlo confidence intervals that explicitly acknowledge the imprecision in the estimated subgroup memberships via perturbations using a parametric bootstrap. The proposal is widely applicable and avoids stringent causal or structural assumptions that existing bias-adjustment or bias-correction methods rely on.Results: Using two different publicly available real-world datasets, we illustrate how the proposed strategy supplements existing latent subgroup effect analyses to shed light on the potential impact of classification uncertainty on inference. First, individuals are partitioned into latent subgroups based on their medical and health history. Then within each fixed latent subgroup, the average treatment effect is assessed using an augmented inverse propensity score weighted estimator. Finally, utilizing the proposed sensitivity analysis reveals different subgroup-specific effects that are mostly insensitive to potential misclassification.Conclusions: Our proposed sensitivity analysis is straightforward to implement, provides both graphical and numerical summaries, and readily permits assessing the sensitivity of any machine learning-based causal effect estimator to classification uncertainty. We recommend making such sensitivity analyses more routine in latent subgroup effect analyses. [ABSTRACT FROM AUTHOR]

Estimation of the diagnostic performance of serological tests often relies on another test assumed as a reference or on samples of known infection status, yet both are seldom available for emerging pathogens in wildlife. Longitudinal disease serological data can be analysed through multi‐event capture–mark–recapture (MECMR) models accounting for the uncertainty in state assignment, allowing us to estimate epidemiological parameters such as incidence and mortality. We hypothesized that by estimating the uncertainty in state assignment, MECMR models estimate the diagnostic performance of serological tests for rabbit haemorrhagic disease virus (RHDV) and myxoma virus (MYXV). We evaluated this hypothesis on longitudinal serological data of three tests of RHDV and one test of MYXV in two populations of the European rabbit (Oryctolagus cuniculus algirus). First, we selected the optimal cut‐off threshold for each test using finite mixture models, a reference method not relying on reference tests or samples. Second, we used MECMR models to compare the diagnostic sensitivity (Se) and specificity (Sp) of the three tests for RHDV. Third, we compared the estimates of diagnostic performance by MECMR and finite mixture models across a range of cut‐off values. The MECMR models showed that the RHDV test employing GI.2 antigens (Se: 100%) outperformed two tests employing GI.1 antigens (Se: 21.7% ± 8.6% and 8.7% ± 5.9%). At their selected cut‐offs (2.0 for RHDV GI.2 and 2.4 for MYXV), the estimates of Se and Sp were concordant between the MECMR and finite mixture models. Over the duration of the study (May 2018 to September 2020), the monthly survival of European rabbits seropositive for MYXV was significantly higher than that of seronegative rabbits (82.7% ± 4.9% versus 61.5% ± 12.7%) at the non‐fenced site. We conclude that MECMR models can reliably estimate the diagnostic performance of serological tests for RHDV and MYXV in European rabbits. This conclusion could extend to other diagnostic tests and host‒pathogen systems. Longitudinal disease surveillance data analysed through MECMR models allow the validation of diagnostic tests for emerging pathogens in novel host species while simultaneously estimating epidemiological parameters. [ABSTRACT FROM AUTHOR]

The paper discusses the specification of finite mixture models based on the Discretized Beta distribution for the analysis of ordered discrete responses, as ratings and count data. The ultimate goal of the paper is to parameterize clusters of opposite and intermediate response outcomes. After a thorough discussion on model interpretation, identifiability and estimation, the proposal is illustrated on the wake of a case study on the probability to vote for German Political Parties and with a comparative discussion with the state of the art. [ABSTRACT FROM AUTHOR]

Robust high dimensional estimation is one of the most important problems in statistics. In a high dimensional structure with a small number of non-zero observations, the dimension of the parameters is larger than the sample size. For modeling the sparsity of outlier response vector, we randomly selected a small number of observations and corrupted them arbitrarily. There are two distinct ways to overcome sparsity in the generalized linear model (GLM): in the parameter space, or in the space output. According to several studies in corrupted observation modeling, there is a relationship between robustness and sparsity. In this paper for obtaining robust high dimensional estimation, we proposed a finite mixture of the generalized linear models (FMGLMs). By using simulation with the expectation-maximization (EM) algorithm, we show improved modeling performance. [ABSTRACT FROM AUTHOR]

Summary: Sufficient dimension reduction (SDR) is an attractive approach to regression modelling. However, despite its rich literature and growing popularity in application, surprisingly little research has been done on how to perform SDR for clustered data, for example as is commonly arises in longitudinal studies. Indeed, current popular SDR methods have been mostly based on a marginal estimating equation approach. In this article, we propose a new approach to SDR for clustered data based on a combination of finite mixture modelling and mixed effects regression. Finite mixture models offer a flexible means of estimating the fixed effects central subspace, based on slicing the space up and probabilistically clustering observations to each slice (mixture component). Dimension reduction is achieved by having the mixing proportions vary only through the sufficient fixed effect predictors. We then incorporate random effects as a natural means of accounting for correlations within clusters. We employ a Monte Carlo expectation–maximisation algorithm to estimate the model parameters and fixed effects central subspace, and discuss methods for associated uncertainty quantification and prediction. Simulation studies demonstrate that our approach performs strongly against both estimating equation methods for estimating the fixed effects central subspace, and SDR methods which do not account for within‐cluster correlation. Finally, we apply the proposed approach to a data set on air pollutant monitoring across 13 stations in the Eastern United States. [ABSTRACT FROM AUTHOR]

Alarm systems are becoming increasingly important in ensuring high levels of safety, lowering pollutants, and reducing the financial losses associated with industrial processes. An industrial alarm system's goal is to detect an unwanted deviation in the process variable (PV) as soon as possible. A thorough understanding of PV behaviour is required when constructing an alarm system. If distinct PV's behaviour features are not taken into account, problems like flooding and chattering will occur. While the PV belongs to a vital industrial site, such as a nuclear plant, the severity of the damage will grow. Statistical techniques are used to design an alarm system in three stages: creating a model for PV based on statistical features, calculating performance assessment indices (FAR, MAR, and AAD) in the presence of design scenarios, and selecting the optimum design policy using optimization algorithms. The first two steps are the topic of this article. To do this, the Finite Mixture Model is utilized to model the behaviour of an Intermittent Fault (IF), which is a fault that alternates between faulty and non-faulty behaviour at discrete random intervals. Finally, different scenarios are explored, and the designed alarm system is tested over time for each of them, with Monte-Carlo simulation being used for further validation. [Display omitted] • A novel method to assess alarm system over the time. • Using time-variant finite mixture models to model the faulty process variables. • Using different scenarios to optimize the alarm system based on the new model. • Evaluating the utilized method by Monte-Carlo simulation. [ABSTRACT FROM AUTHOR]

A general Markov‐Switching autoregressive conditional mean model, valued in the set of non‐negative numbers, is considered. The conditional distribution of this model is a finite mixture of non‐negative distributions whose conditional mean follows a GARCH‐like dynamics with parameters depending on the state of a Markov chain. Three different variants of the model are examined depending on how the lagged‐values of the mixing variable are integrated into the conditional mean equation. The model includes, in particular, Markov mixture versions of various well‐known non‐negative time series models such as the autoregressive conditional duration model, the integer‐valued GARCH (INGARCH) model, and the Beta observation driven model. For the three variants of the model, conditions are given for the existence of a stationary and ergodic solution. The proposed conditions match those already known for Markov‐switching GARCH models. We also give conditions for finite marginal moments. Applications to various mixture and Markov mixture count, duration and proportion models are provided. [ABSTRACT FROM AUTHOR]