Model-based Cluster Analysis Research Articles

Patterns of health care use related to respiratory conditions in early life: A birth cohort study with linked administrative data.

To identify distinctive patterns of respiratory-related health services use (HSU) between birth and 3 years of age, and to examine associated symptom and risk profiles. This study included 729 mother and child pairs enrolled in the Toronto site of the Canadian Healthy Infant Longitudinal Development study in 2009-2012; they were linked to Ontario health administrative databases (2009-2016). A model-based cluster analysis was performed to identify distinct groups of children who followed a similar pattern of respiratory-related HSU between birth and 3 years of age, regarding hospitalization, emergency department (ED) and physician office visits for respiratory conditions and total health care costs (2016 Canadian dollars). The majority (estimated cluster weight = 0.905) showed a pattern of low and stable respiratory care use (low HSU) while the remainder (weight = 0.095) showed a pattern of high use (high HSU). From 0 to 3 years of age, the low- and high-HSU groups differed in mean trajectories of total health care costs ($783 per 6 months decreased to $114, vs $1796 to $177, respectively). Compared to low-HSU, the high-HSU group was associated with a constant risk of hospitalizations, early high ED utilization and physician visits for respiratory problems. The two groups differed significantly in the timing of wheezing (late onset in low-HSU vs early in high-HSU) and future total costs (stable vs increased). One in ten children had high respiratory care use in early life. Such information can help identify high-risk young children in a large population, monitor their long-term health, and inform resource allocation.

Retrotransposon-based genetic variation of Poa annua populations from contrasting climate conditions.

BackgroundPoa annua L. is an example of a plant characterized by abundant, worldwide distribution from polar to equatorial regions. Due to its high plasticity and extraordinary expansiveness, P. annua is considered an invasive species capable of occupying and surviving in a wide range of habitats including pioneer zones, areas intensively transformed by human activities, remote subarctic meadows and even the Antarctic Peninsula region.MethodsIn the present study, we evaluated the utility of inter-primer binding site (iPBS) markers for assessing the genetic variation of P. annua populations representing contrasting environments from the worldwide range of this species. The electrophoretic patterns of polymerase chain reaction products obtained for each individual were used to estimate the genetic diversity and differentiation between populations.ResultsiPBS genotyping revealed a pattern of genetic variation differentiating the six studied P. annua populations characterized by their different climatic conditions. According to the analysis of molecular variance, the greatest genetic variation was recorded among populations, whereas 41.75% was observed between individuals within populations. The results of principal coordinates analysis (PCoA) and model-based clustering analysis showed a clear subdivision of analyzed populations. According to PCoA, populations from Siberia and the Kola Peninsula were the most different from each other and showed the lowest genetic variability. The application of STRUCTURE software confirmed the unique character of the population from the Kola Peninsula.DiscussionThe lowest variability of the Siberia population suggested that it was subjected to genetic drift. However, although demographic expansion was indicated by negative values of Fu’s FS statistic and analysis of mismatch distribution, it was not followed by significant traces of a bottleneck or a founder effect. For the Antarctic population, the observed level of genetic variation was surprisingly high, despite the observed significant traces of bottleneck/founder effect following demographic expansion, and was similar to that observed in populations from Poland and the Balkans. For the Antarctic population, the multiple introduction events from different sources are considered to be responsible for such an observation. Moreover, the results of STRUCTURE and PCoA showed that the P. annua from Antarctica has the highest genetic similarity to populations from Europe.ConclusionsThe observed polymorphism should be considered as a consequence of the joint influence of external abiotic stress and the selection process. Environmental changes, due to their ability to induce transposon activation, lead to the acceleration of evolutionary processes through the production of genetic variability.

Unequal Social Protection under the Federalist System: Three Unemployment Insurance Approaches in the United States, 2007–2015

AbstractDespite the increased attention paid to federal-state unemployment insurance (UI) schemes after the Great Recession (2007–2009), research examining the policy characteristics and the underlying logic shaping the social protection provided by a federal-state UI system is limited. Integrating the perspectives of policy design theory, comparative welfare politics, and fiscal welfare, this paper examines the unequal social protection under the American UI system during and after the Great Recession. By using model-based cluster analysis and fixed-effect panel regression models, this paper identifies three distinct UI approaches, i.e. the limited social protection approach, the unbalanced social protection approach, and the balanced social protection approach. The policy choices made by those states that follow the three approaches reflect different mixtures of policy logic, including social protection, economic stabilization, work disincentives, and interstate competition. The overall downward trend in social protection signals that the American UI system is under-prepared for the next economic recession, thereby exposing unemployed workers to the risk of economic insecurity. These findings provide implications for future policy designs aiming to strengthen the social protection of the federal-state UI system.

Genetic mixture analyses in support of restoration of a high value recreational fishery for rainbow trout (Oncorhynchus mykiss) from a large lake in interior British Columbia

Genetic mixture analysis is an important tool to apportion catch amongst potential component populations contributing to a fishery. We used variation at 10 microsatellite DNA loci to assess the level of genetic divergence between two ecotypes of rainbow trout (Oncorhynchus mykiss) that naturally co-occur in Kootenay Lake, southeastern British Columbia, and to exploit such divergence in a mixture analysis. One form, “Gerrard” rainbow trout, historically matured at sizes greater than 60 cm and 5 kg, spawns at a lake outlet after upstream migration in a large river tributary to Kootenay Lake, and is highly prized in the recreational fishery. The other form, “non-Gerrard” rainbow trout, is also native to the lake and matures at smaller sizes and spawns in numerous small streams tributary to Kootenay Lake. Recent declines in growth rate of Gerrard rainbow trout, however, has made them difficult to identify by size in fishery samples. Gerrard (N = 130, 6 sites) and non-Gerrard trout N = 312, 15 sites) were highly divergent from one another (FST = 0.14, P < 0.001) and constituted distinct genetic groups in model-based clustering analyses. Genetic mixture analyses of fishery samples indicated a high degree of accuracy in estimating mixture proportions; 100% Gerrard simulated fisheries were estimated to contain 99.9% Gerrards (95% confidence intervals of 99.8–100%) while 100% non-Gerrard rainbow trout simulated fisheries were estimated to contain 100.0% non-Gerrard trout (100–100%). Across eight fishery creel samples obtained between 2015 and 2017 (N = 527 fish), mixture analysis estimated the fishery to contain an average of 73.4% (95% confidence interval = 68.4–74.6%) Gerrard and 26.6% (23.4–31.6%) non-Gerrard trout. Realistic fishery simulations demonstrated strong agreement with empirical results; the average simulated values for Gerrards was 73.4% (65.9–80.0%) and for non-Gerrards was 26.8% (20.0–34.1%). Assignment tests resulted in an average 98.5% (± 0.066%) assignment confidence; 385 fish from the fishery samples were assigned to the Gerrard group (0.73) and 142 (0.27) to the non-Gerrard rainbow trout genetic group. Fitting length-at-age data for genetically assigned fishery samples to a von Bertalanffy growth model found greatest support for a model employing ecotype-specific L∞ (= 59.6 and 52.9 cm for Gerrards and non-Gerrards, respectively, both P < 0.001), and t0 (= − 1.36 and − 2.58, respectively, both P < 0.05), but a common K (= 0.189, P < 0.001). Our mixture analyses are being used to monitor catches and better understand the feeding and migration biology of these sympatric ecotypes of rainbow trout.

Identifying Psychopathy Subtypes Using a Broader Model of Personality: An Investigation of the Five Factor Model Using Model-Based Cluster Analysis.

The field of psychopathy research largely argues that the disorder is a heterogeneous concept with two predominant subtypes: primary psychopathy and secondary psychopathy. Early research examined psychopathy subtypes in terms of traits specific to psychopathy. However, recent research suggests that psychopathy may be conceptualized more dimensionally as an extreme variation in normative personality traits. This concept propels the examination of psychopathy using a broader model of personality, the Five Factor Model. Using model-based cluster analysis, this study uses the Five Factor Model to identify subtypes. Utilizing an American college sample, a cluster analysis was performed on the highest third of Psychopathic Personality Inventory (PPI)-Total scorers using the Neuroticism Extraversion Openness Personality Inventory-Revised (NEO-PI-R) domains. This analysis yielded a two-cluster solution, with one cluster representing Primary psychopathy and the other representing Secondary psychopathy. The two clusters displayed differences in Neuroticism, Conscientiousness, Openness, and Extraversion, which may be important for understanding psychopathy subtyping using general personality traits. Associated traits such as aggression, anxiety, and borderline traits largely match other subtyping research. Overall, the results support the existence of psychopathy variants in a nonoffending sample, the dimensionality of psychopathy, and the ability of normative personality traits to differentiate psychopathy subtypes. Unveiling the complex nature of psychopathy through illumination of its dimensions has strong implications for future research as it can help specify an otherwise elusive disorder. Such desired specification can reveal psychopathy's etiology and potential treatment options. The results of this research in particular suggest a link between profiles of normal personality traits and aggressive behavior indicating the potential for risk assessment. Moreover, this research provides evidence for psychopathy's dimensionality versus categorical casting, contributing to the debate about mental disorder classifications in general.

Genetic variation and population structure among larvalLethenteronspp. within the Yukon River drainage, Alaska

The absence of information on genetic variation and population structure of lampreys Lethenteron spp. in the eastern part of their distribution limits our understanding of the migration ecology and spatial population genetic structure of the species. We examined genetic variation within and among three aggregations of Lethenteron spp. larvae in the Yukon River drainage, Alaska, using microsatellite genotypes. A total of 120 larval lampreys were genotyped at eight microsatellite loci. Global FST was 0.053 (95% CI 0.021-0.086), while pairwise FST values ranged from 0.048-0.057. Model-based Bayesian clustering analyses with sample locality priors (LOCPRIOR) identified three distinct, but admixed, genetic clusters that corresponded with the three aggregations. Estimates of contemporary gene flow indicate substantial reciprocal migration among sites consistent with no or low-fidelity natal homing. These results are largely in agreement with previous reports of historic and contemporary gene flow among Lethenteron spp. in other parts of their geographic distribution.

A challenge of predicting seizure frequency in temporal lobe epilepsy using neuroanatomical features

The pathological and clinical characteristics of temporal lobe epilepsy (TLE) tend to be affected by epileptic seizures, specifically represented by seizure lateralization and frequency. Although the lateralization of the epileptogenic zone can be clarified in terms of neuroanatomical damage, there have been conflicting findings on the relationship between seizure frequency and neuroanatomical damage. In this study we sought to examine the relationship in the framework of machine learning-based predictive modeling. We acquired 60 grey matter (GM) anatomical features from structural MRI data and 46 white matter (WM) anatomical features from diffusion-weighted MRI data for 42 TLE patients and 45 healthy controls and applied the random forests method to the neuroanatomical features. We demonstrated that, whereas the neuroanatomical features were promising markers for the discrimination of the TLE patients from the healthy controls, the separation between the TLE patients with low and high seizure frequency on the basis of the neuroanatomical features was challenging. When we applied feature selection techniques for the construction of the predictive models, a greater number of features were selected as being relevant to the distinction of the TLE patients from the healthy controls than to the classification of the TLE patients according to seizure frequency. Furthermore, we adopted model-based clustering analysis and showed that seizure frequency-based subgroups were not matched well with neuroanatomy-based subgroups in the TLE patients. We propose that the challenge of predicting seizure frequency using neuroanatomical features could be attributable to considerable inter-individual variability in neuroanatomical damage among seizure frequency-based subgroups.

Genome-wide regulatory gene-derived SSRs reveal genetic differentiation and population structure in fiber flax genotypes

We designed a set of 580 simple sequence repeat markers; 506 from transcription factor-coding genes, and 74 from long non-coding RNAs and designated them as regulatory gene-derived simple sequence repeat (ReG-SSR) markers. From this set, we could anchor 559 ReG-SSR markers on 15 flax chromosomes with an average marker distance of 0.56Mb. Thirty-one polymorphic ReG-SSR primers, amplifying SSR loci length of at least 20bp were chosen from 134 screened primers. This primer set was used to characterize a diversity panel of 93 flax accessions. The panel included 33 accessions from India, including released varieties, dual-purpose lines and landraces, and 60 fiber flax accessions from the global core collection. Thirty-one ReG-SSR markers generated 76 alleles, with an average of 2.5 alleles per primer and a mean allele frequency of 0.77. These markers recorded 0.32 average gene diversity, 0.26 polymorphism information content and 1.35% null alleles. All the 31 ReG-SSR loci were found selectively neutral and showed no evidence of population reduction. A model-based clustering analysis separated the flax accessions into two sub-populations-Indian and global, with some accessions showing admixtures. The distinct clustering pattern of the Indian accessions compared to the global accessions, conforms to the principal coordinate analysis, genetic dissimilarity-based unweighted neighbor-joining tree and analysis of molecular variance. Fourteen flax accessions with 99.3% allelic richness were found optimum to adopt in breeding programs. In summary, the genome-wide ReG-SSR markers will serve as a functional marker resource for genetic and phenotypic relationship studies, marker-assisted selections, and provide a basis for selection of accessions from the Indian and global gene pool in fiber flax breeding programs.

Trajectories of childhood immune development and respiratory health relevant to asthma and allergy.

Events in early life contribute to subsequent risk of asthma; however, the causes and trajectories of childhood wheeze are heterogeneous and do not always result in asthma. Similarly, not all atopic individuals develop wheeze, and vice versa. The reasons for these differences are unclear. Using unsupervised model-based cluster analysis, we identified latent clusters within a prospective birth cohort with deep immunological and respiratory phenotyping. We characterised each cluster in terms of immunological profile and disease risk, and replicated our results in external cohorts from the UK and USA. We discovered three distinct trajectories, one of which is a high-risk 'atopic' cluster with increased propensity for allergic diseases throughout childhood. Atopy contributes varyingly to later wheeze depending on cluster membership. Our findings demonstrate the utility of unsupervised analysis in elucidating heterogeneity in asthma pathogenesis and provide a foundation for improving management and prevention of childhood asthma.

Complex fine-scale spatial genetic structure in Epidendrum rhopalostele: an epiphytic orchid.

Orchid seeds are presumably dispersed by wind due to their very small size and thus can potentially travel long distances. However, the few related studies indicate that seeds fall close to their mother plants. Because seed dispersal and colonization patterns can have relevant consequences for long-term species persistence, we assessed the fine-scale genetic structure of the epiphytic orchid Epidendrum rhopalostele to provide insight into these patterns. All individuals in the studied population were georeferenced and genotyped with AFLP-markers. Genetic structure was evaluated at two levels (forest and tree) using three approaches: principal coordinates analysis, model-based clustering, and spatial autocorrelation analysis. Results showed two genetic groups, composed of individuals from almost every tree with orchids. Spatial autocorrelation analysis at the forest level found no significant genetic structure when all individuals were considered, but a pattern of genetic patches was revealed when the analysis was performed separately for each group. Genetic patches had an estimated diameter of 4 m and were composed of individuals from more than one tree. A weak genetic structure was detected at the tree level at distances less than 1.5 m. These results suggest that many seeds fall close to the mother plant and become established in the same host tree. Additionally, a sequential colonization process seems to be the predominant mode of expansion, whereby progeny from orchids in one tree colonize neighboring trees. Thus, the existence of two distinct genetic groups and the presence of genetic patches should be considered when seed sampling for ex situ conservation.

Twigs on the same branch? Identifying personality profiles in police officers using psychopathic personality traits

Recent high-profile incidents reignited the conversation about psychopathic traits in police officers. Psychopathy is characterized by multiple variants: primary and secondary psychopathy. There is limited research examining psychopathy in populations that may exhibit adaptive psychopathic traits. This study used model-based cluster analyses of high psychopathy scorers to investigate psychopathic subtypes in an urban police sample. Relative to the primary subtype, the secondary group displayed higher levels of Self-Centered Impulsivity, trait anxiety, covert narcissism, borderline personality disorder traits, substance use, psychiatric treatment, and aggression. These findings support the concept of successful psychopathy and the existence of psychopathy profiles in police officers, providing a useful look at how successful psychopathy may manifest as well as implications for the criminal justice system and police departments.

Genetic flow among olive populations within the Mediterranean basin.

BackgroundThe olive tree is a typical crop of the Mediterranean basin where it shows a wide diversity, accounting for more than 2,600 cultivars. The ability to discriminate olive cultivars and determine their genetic variability is pivotal for an optimal exploitation of olive genetic resources.MethodsWe investigated the genetic diversity within 128 olive accessions belonging to four countries in the Mediterranean Basin (Italy, Algeria, Syria, and Malta), with the purpose of better understanding the origin and spread of the olive genotypes across Mediterranean Basin countries. Eleven highly polymorphic simple sequence repeat (SSR) markers were used and proved to be very informative, producing a total of 179 alleles.ResultsCluster analysis distinguished three main groups according to their geographical origin, with the current sample of Maltese accessions included in the Italian group. Phylogenetic analysis further differentiated Italian and Maltese olive accessions, clarifying the intermediate position of Maltese accessions along the x/y-axes of principal coordinate analysis (PCoA). Model-based and neighbor clustering, PCoA, and migration analysis suggested the existence of two different gene pools (Algerian and Syrian) and that the genetic exchange occurred between the Syrian, Italian and Maltese populations.DiscussionThe close relationship between Syrian and Italian and Maltese olives was consistent with the historical domestication and migration of olive tree from the North Levant to eastern Mediterranean basin. This study lays the foundations for a better understanding of olive genetic diversity in the Mediterranean basin and represents a step toward an optimal conservation and exploitation of olive genetic resources.

A study on population dynamics in \u201cBelt & Road\u201d countries and their implications

China’s Belt & Road (“B&R”) initiative has attracted much attention in recent years. Many studies have assessed the potential impacts of this initiative from the economic, political, foreign relations, and military perspectives. However, very little attention has been paid to examine the opportunities and challenges of this new initiative from a demographic perspective. Therefore, this study investigates into the population dynamics of 65 countries in the B&R region, and explores the demographic dynamics of this regional cooperation model. Based on graphic visualization, the demographic heterogeneities within the B&R region, in terms of current stages of demographic transitions, population age structure, and demographic windows of opportunities have been revealed. Furthermore, a model-based cluster analysis has been conducted to classify the B&R countries into two groups. One group consists of countries with younger population, low education access, and larger gender gap in the labor market, while the other group comprises of more aged population and has better performances in economic development, education enrolment, and female labor participation. The demographic windows of opportunities in the former group will remain open for the coming decades; whereas the windows have already closed or is about to close very soon in the latter group. The former group should take lessons from the latter concerning how to better prepare themselves for the demographic transition. Moreover, the curvilinear relationship between wealth and well-being has a complicated structure and the B&R initiative should focus not only on economic development but also on the human well-being.

Salience of primary and secondary school students' achievement emotions and perceived antecedents: Interviews on literacy and mathematics domains

Within the theoretical framework provided by Pekrun's (2006) control-value theory, we investigated individual differences in students' achievement emotions and antecedents. The subjects were 149 fourth, seventh, and eleventh-graders, who participated in semi-structured interviews focused on how they usually felt when facing literacy or mathematics, and on their perception of the underlying antecedents that triggered the emotions. Generalized Linear Mixed Models (GLMM) indicated higher salience of enjoyment and relaxation compared to a range of other positive and negative emotions. There were class level differences in terms of a higher number of spontaneously reported emotions for eleventh-graders compared to younger students. Through the model-based cluster analysis we identified three groups of students, each characterized by a different pattern of emotions, not differing by class level, domain, or gender. Applying GLMM revealed some differences between groups in the pattern of perceived antecedents. Results are discussed considering their theoretical and applied relevance.

Differential outcome subgroups in children with autism spectrum disorder attending early intervention.

The finding of positive outcomes at the group level for children with autism spectrum disorder (ASD) who complete comprehensive early intervention programmes often masks considerable individual variability. We therefore aimed to identify subgroups of children based on their response to intervention and to compare outcome variables between groups at two points in time. We used model-based cluster analysis to explore response to intervention using a longitudinal design for 210 children with ASD who had completed an early intervention programme. Children were assessed on entry at time 1 and again at time 2, which was after 12months or when they exited the programme (whichever came first) using measures of ASD symptoms (Social Communication Questionnaire), cognition (Mullen Scales of Early Learning) and adaptive behaviour (Vineland Adaptive Behaviour Scales-II). A two-cluster solution was identified, including a high change group who improved consistently more than the low change group across measures, and showed significantly fewer autism symptoms, higher non-verbal and verbal cognition and adaptive behaviour composite scores at time 1. The findings indicated that children's response to early intervention is not uniform but instead included subgroups characterised by patterns of high and low change. Further research is needed to identify clinically relevant mediators of differential response group membership.

Personality Traits Differentiate Subgroups of Criminal Offenders With Distinct Cognitive, Affective, and Behavioral Profiles

Mental illness and substance use disorder are prominent among jail and prison inmates. Developing more effective mental and behavioral health treatment approaches for incarcerated criminal offenders is a critical step toward reducing rates of recidivism and relapse following release. Specifying subgroups of offenders who differ in cognitive, affective, behavioral, and demographic characteristics could facilitate more targeted and effective treatment approaches. Accordingly, this study utilized model-based cluster analysis to identify subgroups of offenders in a large sample of incarcerated male offenders ( n = 2,388) using broad personality traits. Six subgroups of offenders emerged based on personality scales related to positive affect, negative affect, and behavioral control. Subsequent analyses revealed significant differences between groups on measures of cognitive ability, affect dysregulation, antisociality, substance use, and impulsivity. These findings provide insight into potential treatment and intervention strategies to target the specific needs of distinct subgroups of incarcerated offenders.

Metrics and methods for characterizing dairy farm intensification using farm survey data.

Evaluation of agricultural intensification requires comprehensive analysis of trends in farm performance across physical and socio-economic aspects, which may diverge across farm types. Typical reporting of economic indicators at sectorial or the “average farm” level does not represent farm diversity and provides limited insight into the sustainability of specific intensification pathways. Using farm business data from a total of 7281 farm survey observations of English and Welsh dairy farms over a 14-year period we calculate a time series of 16 key performance indicators (KPIs) pertinent to farm structure, environmental and socio-economic aspects of sustainability. We then apply principle component analysis and model-based clustering analysis to identify statistically the number of distinct dairy farm typologies for each year of study, and link these clusters through time using multidimensional scaling. Between 2001 and 2014, dairy farms have largely consolidated and specialized into two distinct clusters: more extensive farms relying predominantly on grass, with lower milk yields but higher labour intensity, and more intensive farms producing more milk per cow with more concentrate and more maize, but lower labour intensity. There is some indication that these clusters are converging as the extensive cluster is intensifying slightly faster than the intensive cluster, in terms of milk yield per cow and use of concentrate feed. In 2014, annual milk yields were 6,835 and 7,500 l/cow for extensive and intensive farm types, respectively, whilst annual concentrate feed use was 1.3 and 1.5 tonnes per cow. For several KPIs such as milk yield the mean trend across all farms differed substantially from the extensive and intensive typologies mean. The indicators and analysis methodology developed allows identification of distinct farm types and industry trends using readily available survey data. The identified groups allow the accurate evaluation of the consequences of the reduction in dairy farm numbers and intensification at national and international scales.

Three Worlds of Representation: A Linkage-Based Typology of Parties in Western and Eastern Europe

This article is part of the special cluster titled Parties and Democratic Linkage in Post-Communist Europe, guest edited by Lori Thorlakson, and will be published in the August 2018 issue of EEPS Have Eastern European democracies developed patterns of accountability similar to those existent in their established counterparts? While most accounts of convergence are confined to the world of programmatic reasoning and policy representation, we use a unique data set to cover the wealth of instrumental and emotional modes of linkage building. We apply advanced techniques of model-based cluster analysis to establish a linkage-based typology of political parties. In the East, the contrast of programmatic and clientelistic parties is most essential in the absence of strong regional subdivisions. In the West, the structure of linkage building is characterized by an all-encompassing divide that separates mainstream and challenger parties. Parties in Southern Europe form a distinct Mediterranean type of “machine politics.” The results for affluent post-industrial societies both support and contradict premises of the cartel party hypothesis. For third-wave democracies in the East, our results suggest the persistence of legacies of pre-communist and communist rule against the weight of cumulative democratic experience. In sum, patterns of accountability remained markedly different in the two regions on the eve of the economic crisis in 2008–2009.

Variable selection in model-based clustering and discriminant analysis with a regularization approach

Several methods for variable selection have been proposed in model-based clustering and classification. These make use of backward or forward procedures to define the roles of the variables. Unfortunately, such stepwise procedures are slow and the resulting algorithms inefficient when analyzing large data sets with many variables. In this paper, we propose an alternative regularization approach for variable selection in model-based clustering and classification. In our approach the variables are first ranked using a lasso-like procedure in order to avoid slow stepwise algorithms. Thus, the variable selection methodology of Maugis et al. (Comput Stat Data Anal 53:3872–3882, 2009b) can be efficiently applied to high-dimensional data sets.

The Party Structure of Mutual Funds

We investigate the structure of mutual funds’ corporate governance preferences as revealed by how they vote their shares in portfolio companies. We apply unsupervised learning tools from the machine learning literature to analyze a comprehensive dataset of mutual funds’ votes and find that a parsimonious two-dimensional model can explain the bulk of mutual fund voting. The two dimensions capture competing visions of corporate governance and are closely related to the recommendations of the leading proxy advisors. Model-based cluster analysis shows that mutual funds are organized into three ‘parties’—the Traditional Governance Party, the Shareholder Reform Party, and the Shareholder Protest Party—that follow distinctive philosophies of corporate governance and shareholders’ role. Our preference measures for mutual funds generate a range of insights about the broader system of corporate governance.