Posterior Membership Probabilities Research Articles

Longitudinal latent polysubstance use patterns among a cohort of people who use opioids in Vancouver, Canada.

Polysubstance use (PSU) practices are increasing among people who use opioids (PWUO). However, several aspects of longitudinal PSU patterns among PWUO remain understudied. This study aims to identify person-centred longitudinal patterns of PSU among a cohort of PWUO. Using longitudinal data (2005-2018) from three prospective cohort studies including people who use drugs in Vancouver, Canada, we used repeated measures latent class analysis to identify different PSU classes among PWUO. Multivariable generalised estimating equations models weighted by the respective posterior membership probabilities were applied to identify covariates of membership in different PSU classes over time. Overall, 2627 PWUO (median age at baseline: 36 [quartile 1-3: 25-45]) were included between 2005 and 2018. We found five distinct PSU patterns, including low/infrequent probability of regular substance use (Class 1; 30%), primarily opioid and methamphetamine use (Class 2; 22%), primarily cannabis use (Class 3; 15%), primarily opioid and crack use (Class 4; 29%) and frequent PSU (Class 5; 4%). Membership in Class 2, 4 and 5 was positively associated with several behavioural and socio-structural adversities. Findings of this longitudinal study suggest PSU is the norm among PWUO and highlights the heterogeneous characteristics of PWUO. The diversities within the population of PWUO need to be recognised in addiction care and treatment as well as optimising resource allocation in the response to the overdose crisis.

Population Preferences for Primary Care Models for Hypertension in Karnataka, India

Hypertension contributes to more than 1.6 million deaths annually in India, with many individuals being unaware they have the condition or receiving inadequate treatment. Policy initiatives to strengthen disease detection and management through primary care services in India are not currently informed by population preferences. To quantify population preferences for attributes of public primary care services for hypertension. This cross-sectional study involved administration of a household survey to a population-based sample of adults with hypertension in the Bengaluru Nagara district (Bengaluru City; urban setting) and the Kolar district (rural setting) in the state of Karnataka, India, from June 22 to July 27, 2021. A discrete choice experiment was designed in which participants selected preferred primary care clinic attributes from hypothetical alternatives. Eligible participants were 30 years or older with a previous diagnosis of hypertension or with measured diastolic blood pressure of 90 mm Hg or higher or systolic blood pressure of 140 mm Hg or higher. A total of 1422 of 1927 individuals (73.8%) consented to receive initial screening, and 1150 (80.9%) were eligible for participation, with 1085 (94.3%) of those eligible completing the survey. Relative preference for health care service attributes and preference class derived from respondents selecting a preferred clinic scenario from 8 sets of hypothetical comparisons based on wait time, staff courtesy, clinician type, carefulness of clinical assessment, and availability of free medication. Among 1085 adult respondents with hypertension, the mean (SD) age was 54.4 (11.2) years; 573 participants (52.8%) identified as female, and 918 (84.6%) had a previous diagnosis of hypertension. Overall preferences were for careful clinical assessment and consistent availability of free medication; 3 of 5 latent classes prioritized 1 or both of these attributes, accounting for 85.1% of all respondents. However, the largest class (52.4% of respondents) had weak preferences distributed across all attributes (largest relative utility for careful clinical assessment: β = 0.13; 95% CI, 0.06-0.20; 36.4% preference share). Two small classes had strong preferences; 1 class (5.4% of respondents) prioritized shorter wait time (85.1% preference share; utility, β = -3.04; 95% CI, -4.94 to -1.14); the posterior probability of membership in this class was higher among urban vs rural respondents (mean [SD], 0.09 [0.26] vs 0.02 [0.13]). The other class (9.5% of respondents) prioritized seeing a physician (the term doctor was used in the survey) rather than a nurse (66.2% preference share; utility, β = 4.01; 95% CI, 2.76-5.25); the posterior probability of membership in this class was greater among rural vs urban respondents (mean [SD], 0.17 [0.35] vs 0.02 [0.10]). In this study, stated population preferences suggested that consistent medication availability and quality of clinical assessment should be prioritized in primary care services in Karnataka, India. The heterogeneity observed in population preferences supports considering additional models of care, such as fast-track medication dispensing to reduce wait times in urban settings and physician-led services in rural areas.

Distinct Latent Trajectory Phenotypes of Left Ventricular Function Associated with Differential Survival Among Children Treated for Acute Myeloid Leukemia: A Report from the Children's Oncology Group

Background: Children treated for acute myeloid leukemia (AML) receive high doses of anthracyclines that may cause substantial short- and long-term morbidity and mortality. Thus, understanding its natural history and impact on treatment outcomes is crucial. Prior evaluations of cardiotoxicity in pediatric AML have focused on occurrence of incident declines below a left ventricular (LV) ejection fraction (EF) threshold (e.g., EF <50%). Broadening the focus to longitudinal EF trajectories may provide a mechanism to better characterize high-risk phenotypes, inform prognosis, and identify critical windows for intervention. Methods: Between June 2011 and September 2018, AAML1031 enrolled patients aged <30 years with de novo AML. Echocardiographic evaluations were required before each treatment course, at the end of protocol therapy, and annually during off-protocol follow-up. The resulting EF measures were prospectively collected. Group-based trajectory modeling was used to identify latent subgroups with distinct patterns of longitudinal change in EF from baseline. Analyses were restricted to patients with a baseline measure of EF and at least two additional EF measures over the study period. The maximum number of potential EF trajectory classes was limited a priori to five. First, a single quadratic trajectory was modeled, and its fit statistics were compared to 2-, 3-, 4- and 5-group models. Then, the shapes of the changes in EF over time were modeled comparing fit statistics for models consisting of all combinations of linear, quadratic, and cubic trajectories. Individual patients were assigned to a single trajectory group based on their maximum posterior probability of membership. Average posterior probabilities of group membership were computed for each identified latent trajectory phenotype; average posterior probabilities >0.70 indicate that the model adequately discriminates between individuals with different trajectories. Distributions of demographic and clinical characteristics were compared for trajectory groups identified by the final model. Multivariable adjusted Cox regressions compared event-free (EFS) and overall survival (OS) for the identified EF trajectory groups. Analyses restricted to patients who completed all planned therapy on AAML1031 were also performed. Results: 1063 patients contributed a total of 7284 EF measurements (median 6, IQR: 5-11 per person). The final model identified three distinct latent groups, one linear and two cubic trajectories (Figure 1), with 47% of patients categorized into Group 1 (preserved EF), 47% into Group 2 (moderate persistent reduction in EF), and 6% into Group 3 (dramatic EF decline with incomplete recovery). Average posterior probabilities were 0.87, 0.86, and 0.90 for Groups 1-3, respectively. The proportion of infants decreased from Group 1 to Group 3 (G1: 23%, G2: 16%, G3:11%; p<0.01). Group 3 had a higher proportion of Non-Hispanic Black patients (G1: 12%, G2: 11%, G3: 24% v, p=0.02) and slightly higher proportions of females (G1: 49%, G2: 46%, G3: 58%; p=0.17) compared to Groups 1 and 2. Groups 2 and 3 were less likely to have received dexrazoxane compared to Group 1 (G1: 17%, G2: 7%, G3: 7% v, p<0.001). There were no significant differences in distributions of insurance, weight status, risk classification or treatment arm. Group 2 EFS (G1: 48%, G2: 47%; p=0.72) and OS (G1: 68%, G2:64% v, p=0.22) were comparable to Group 1 whereas Group 3 EFS (G1: 48%, G3: 30%; p=0.003) and OS (G1: 68%, G3: 38%; p<0.001) were significantly reduced (Figure 2). Analyses restricted to patients completing AAML1031 therapy identified comparable trajectories but with higher average posterior probabilities (G1: 0.89, G2: 0.90, and G3: 0.94) as these patients had more complete capture of LVEF measurements in the study database.Conclusions: Group-based trajectory methods identified distinct EF trajectory phenotypes with differential EFS and OS. Prediction models focused on trajectory-defined phenotypes may elucidate genomic and environmental determinants of anthracycline-associated cardiotoxicity more effectively than investigations of incident occurrence alone. Accurate prediction of EF trajectory phenotypes may inform personalization of prognosis, treatment, and echo monitoring for pediatric patients treated for AML. Work is ongoing to identify supportive care practices that may modify distinct EF trajectory phenotypes. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

Item-Fit Statistic Based on Posterior Probabilities of Membership in Ability Groups.

A novel approach to item-fit analysis based on an asymptotic test is proposed. The new test statistic, , compares pseudo-observed and expected item mean scores over a set of ability bins. The item mean scores are computed as weighted means with weights based on test-takers' a posteriori density of ability within the bin. This article explores the properties of in case of dichotomously scored items for unidimensional IRT models. Monte Carlo experiments were conducted to analyze the performance of . Type I error of was acceptably close to the nominal level and it had greater power than Orlando and Thissen's . Under some conditions, power of also exceeded the one reported for the computationally more demanding Stone's .

Posterior probabilities of membership of repertoires in acoustic clades.

Recordings of calls may be used to assess population structure for acoustic species. This can be particularly effective if there are identity calls, produced nearly exclusively by just one population segment. The identity call method, IDcall, classifies calls into types using contaminated mixture models, and then clusters repertoires of calls into identity clades (potential population segments) using identity calls that are characteristic of the repertoires in each identity clade. We show how to calculate the Bayesian posterior probabilities that each repertoire is a member of each identity clade, and display this information as a stacked bar graph. This methodology (IDcallPP) is introduced using the output of IDcall but could easily be adapted to estimate posterior probabilities of clade membership when acoustic clades are delineated using other methods. This output is similar to that of the STRUCTURE software which uses molecular genetic data to assess population structure and has become a standard in conservation genetics. The technique introduced here should be a valuable asset to those who use acoustic data to address evolution, ecology, or conservation, and creates a methodological and conceptual bridge between geneticists and acousticians who aim to assess population structure.

Abstract 12549: Population Patterns of Clustering in Cardiovascular Health Metrics and Associations With Cardiovacular Disease: A Latent Classpopulation Patterns of Clustering in Cardiovascular Health Metrics and Associations With Cardiovacular Disease: A Latent Class Analysis

Background: Cardiovascular health (CVH) is defined by 7 health behaviors (smoking, diet, physical activity) and factors (body mass index, blood pressure, total cholesterol, glucose). we sought to explore how CVH components cluster in individuals and examine associations of clusters with cardiovascular disease (CVD). Methods: Participants aged 20-79 years were included from four longitudinal cohorts. Levels of each CVH metric were classified as ideal, intermediate, or poor using AHA definitions. Among sex and race subgroup, we applied latent class analysis (LCA) models on CVH metrics and individuals were assigned to a class based on the highest posterior probability of membership. Cox regression model was conducted to estimate the association of cluster and the CVD events. Results: Among 9270 White and 3913 Black women, and 7773 White and 2875 Black men, 6004 CVD events occurred during 19.5 years’ follow-up. Three common classes were observed (Figure 1): Suboptimal (unfavorable behaviors and factors), Mixed (diverse levels of behaviors and factors), and Favorable factors. One additional class of Unfavorable Behaviors (with relatively favorable factors) was identified in Whites. Compared with Favorable factors, other classes had significantly higher risk for CVD (Figure 2), except in Black women. Conclusions: We observed three common clusters of CVH component, with an additional class of Unfavorable Behaviors (only) in White participants, that were significantly associated with incident CVD across subgroups. Given that 27-47% of participants were in the Suboptimal group, multifaceted primordial prevention strategies will be needed to bend the CVD epidemic curve.

PHENOTYPES OF CARDIOVASCULAR DAMAGE AND THE RELATIONSHIP WITH COGNITIVE DECLINE: LATENT CLASS ANALYSIS

The association between subclinical cardiovascular disease (SCVD) and cognitive function is well established, however, current tools for addressing subgroups of SCVD have focused on the overall burden of disease. Identifying risky combinations of characteristics may lead to a better understanding of pathophysiologic changes that underlie cognitive decline. Participants included 5,072 older adults from the Cardiovascular Health Study without cardiovascular disease at baseline and followed for 6 years. Using latent class analysis, we identified cardiovascular damage phenotypes based on vascular (internal intima-media thickness, ankle-arm index, white matter grade, and brain infarctions), cardiac (major echocardiogram abnormalities, ST2, and N-terminal pro-brain natriuretic peptide) and inflammatory (interleukin-6, galectin-3, and cystatin C) markers. Using the maximum probability assignment rule, participants were assigned to phenotypes based on the highest posterior probability of membership. We used adjusted linear mixed effects models to evaluate the association between phenotype and cognitive decline, measured annually using the Modified Mini-Mental State Exam. The analysis yielded 5 prevalent phenotypes: healthy (65%), cardiac (10%), inflammatory (10%), multisystem morbidity (11%), and vascular (4%). The vascular phenotype had the greatest rate of decline at 0.88 points per year (95% CI= -1.33, -0.44), followed by the multisystem morbidity phenotype (β=-0.72, 95% CI= -1.07, -0.38), the inflammatory phenotype (β=-0.67, 95% CI= -0.95, -0.38), and the cardiac phenotype (β=-0.45, 95% CI= -0.70, -0.21) compared to the healthy phenotype. Among patterns of cardiovascular damage, vascular damage appears to have the greatest influence on cognitive decline. Identification of these phenotypes may give rise to more informed clinical interventions.

Trajectories of prescription opioids filled over time.

We performed a retrospective cohort study that aimed to identify one or more groups that followed a pattern of chronic, high prescription use and quantify individuals’ time-dependent probabilities of belonging to a high-utilizer group. We analyzed data from 52,456 adults age 18–45 who enrolled in Medicaid from 2009–2017 in Allegheny County, Pennsylvania who filled at least one prescription for an opioid analgesic. We used group-based trajectory modeling to identify groups of individuals with distinct patterns of prescription opioid use over time. We found the population to be comprised of three distinct trajectory groups. The first group comprised 83% of the population and filled few, if any, opioid prescriptions after their index prescription. The second group (12%) initially filled an average of one prescription per month, but declined over two years to near-zero. The third group (6%) demonstrated sustained high opioid prescriptions utilization. Using individual patients’ posterior probability of membership in the high utilization group, which can be updated iteratively over time as new information become available, we defined a sensitive threshold predictive of sustained future opioid utilization. We conclude that individuals at risk of sustained opioid utilization can be identified early in their clinical course from limited observational data.

Categorical latent variable modeling utilizing fuzzy clustering generalized structured component analysis as an alternative to latent class analysis

Latent class analysis is becoming popular in many areas of education, psychology, social and behavioral sciences, public health, and medicine. However, it often suffers from identification issues due to the large number of parameters involved when using maximum likelihood (ML) estimation. Increasing the sample size, reducing sparseness, and strengthening the relationship between the observed variables and the latent variables all improve the information and thus reduce the identification issues, but the identification issue still affects the validity of parameter estimates in ML estimation and the definition of identification is not sufficient to guarantee the existence of an ML solution. In this paper, generalized structured component analysis (GSCA), which is a component-based approach that utilizes optimal scaling and fuzzy clustering, is applied to avoid these identification issues and develop more stable solutions for the heterogeneity of a population based on a set of categorical responses. Testing our proposed new approach, component-based (CB) latent class analysis (LCA), on real world substance use data from Add Health produced not only the same features as those yielded by conventional ML LCA but also stable estimation without identification issues. Comparing the results obtained from ML LCA using Mplus and poLCA in R, with those from our proposed CB LCA using GSCA in R revealed a similar number of latent classes and posterior probabilities and only minor discrepancies in individual latent class classifications when the posterior probabilities of membership are not distinct.

Cluster analysis for repeated data with dropout: Sensitivity analysis using a distal event

ABSTRACTDegeneration of the aortic wall becomes life-threatening when the risk of rupture increases. Cluster analysis on repeated measures of the diameter of the artery revealed two subgroups of patients included in a surveillance program. These results were obtained under the assumption of missingness at random. In this article, we study the vulnerability of the cluster analysis results – the estimated trajectories and the posterior membership probabilities – by applying different missing-data models for non-ignorable dropout, as proposed by Muthen et al. (2011) to the growth of the diameter of the artery.

The effects of rurality on substance use disorder diagnosis: A multiple-groups latent class analysis

BackgroundRates of accidental overdose mortality from substance use disorder (SUD) have risen dramatically in the United States since 1990. Between 1999 and 2004 alone rates increased 62% nationwide, with rural overdose mortality increasing at a rate 3 times that seen in urban populations. Cultural differences between rural and urban populations (e.g., educational attainment, unemployment rates, social characteristics, etc.) affect the nature of SUD, leading to disparate risk of overdose across these communities. MethodsMultiple-groups latent class analysis with covariates was applied to data from the 2011 and 2012 National Survey on Drug Use and Health (n=12.140) to examine potential differences in latent classifications of SUD between rural and urban adult (aged 18years and older) populations. Nine drug categories were used to identify latent classes of SUD defined by probability of diagnosis within these categories. Once the class structures were established for rural and urban samples, posterior membership probabilities were entered into a multinomial regression analysis of socio-demographic predictors' association with the likelihood of SUD latent class membership. ResultsLatent class structures differed across the sub-groups, with the rural sample fitting a 3-class structure (Bootstrap Likelihood Ratio Test P value=0.03) and the urban fitting a 6-class model (Bootstrap Likelihood Ratio Test P value<0.0001). Overall the rural class structure exhibited less diversity in class structure and lower prevalence of SUD in multiple drug categories (e.g. cocaine, hallucinogens, and stimulants). ConclusionsThis result supports the hypothesis that different underlying elements exist in the two populations that affect SUD patterns, and thus can inform the development of surveillance instruments, clinical services, and prevention programming tailored to specific communities.

Multiple Chronic Conditions and Hospitalizations Among Recipients of Long-Term Services and Supports.

Among older adults receiving long-term services and supports (LTSS), debilitating hospitalizations is a pervasive clinical and research problem. Multiple chronic conditions (MCCs) are prevalent in LTSS recipients. However, the combination of MCCs and diseases associated with hospitalizations of LTSS recipients is unclear. The purpose of this analysis was to determine the association between classes of MCCs in newly enrolled LTSS recipients and the number of hospitalizations over a 1-year period following enrollment. This report is based on secondary analysis of extant data from a longitudinal cohort study of 470 new recipients of LTSS, 60 years and older, receiving services in assisted living facilities, nursing homes, or through home- and community-based services. Using baseline chronic conditions reported in medical records, latent class analysis was used to identify classes of MCCs and posterior probabilities of membership in each class. Poisson regressions were used to estimate the relative ratio between posterior probabilities of class membership and number of hospitalizations during the 3-month period prior to the start of LTSS (baseline) and then every 3 months forward through 12 months. Three latent MCC-based classes named Cardiopulmonary, Cerebrovascular/Paralysis, and All Other Conditions were identified. The Cardiopulmonary class was associated with elevated numbers of hospitalizations compared to the All Other Conditions class (relative ratio [RR] = 1.88, 95% CI [1.33, 2.65], p < .001). Older LTSS recipients with a combination of MCCs that includes cardiopulmonary conditions have increased risk for hospitalization.

A method of real-time fault diagnosis for power transformers based on vibration analysis

In this paper, a novel probability-based classification model is proposed for real-time fault detection of power transformers. First, the transformer vibration principle is introduced, and two effective feature extraction techniques are presented. Next, the details of the classification model based on support vector machine (SVM) are shown. The model also includes a binary decision tree (BDT) which divides transformers into different classes according to health state. The trained model produces posterior probabilities of membership to each predefined class for a tested vibration sample. During the experiments, the vibrations of transformers under different conditions are acquired, and the corresponding feature vectors are used to train the SVM classifiers. The effectiveness of this model is illustrated experimentally on typical in-service transformers. The consistency between the results of the proposed model and the actual condition of the test transformers indicates that the model can be used as a reliable method for transformer fault detection.

Quantitative forecasting of PTSD from early trauma responses: A Machine Learning application

There is broad interest in predicting the clinical course of mental disorders from early, multimodal clinical and biological information. Current computational models, however, constitute a significant barrier to realizing this goal. The early identification of trauma survivors at risk of post-traumatic stress disorder (PTSD) is plausible given the disorder's salient onset and the abundance of putative biological and clinical risk indicators. This work evaluates the ability of Machine Learning (ML) forecasting approaches to identify and integrate a panel of unique predictive characteristics and determine their accuracy in forecasting non-remitting PTSD from information collected within10 days of a traumatic event. Data on event characteristics, emergency department observations, and early symptoms were collected in 957 trauma survivors, followed for fifteen months. An ML feature selection algorithm identified a set of predictors that rendered all others redundant. Support Vector Machines (SVMs) as well as other ML classification algorithms were used to evaluate the forecasting accuracy of i) ML selected features, ii) all available features without selection, and iii) Acute Stress Disorder (ASD) symptoms alone. SVM also compared the prediction of a) PTSD diagnostic status at 15 months to b) posterior probability of membership in an empirically derived non-remitting PTSD symptom trajectory. Results are expressed as mean Area Under Receiver Operating Characteristics Curve (AUC). The feature selection algorithm identified 16 predictors, present in ≥95% cross-validation trials. The accuracy of predicting non-remitting PTSD from that set (AUC = .77) did not differ from predicting from all available information (AUC = .78). Predicting from ASD symptoms was not better then chance (AUC = .60). The prediction of PTSD status was less accurate than that of membership in a non-remitting trajectory (AUC = .71). ML methods may fill a critical gap in forecasting PTSD. The ability to identify and integrate unique risk indicators makes this a promising approach for developing algorithms that infer probabilistic risk of chronic posttraumatic stress psychopathology based on complex sources of biological, psychological, and social information.

Cluster membership probabilities from proper motions and multi-wavelength photometric catalogues

We present a new technique designed to take full advantage of the high dimensionality (photometric, astrometric, temporal) of the DANCe survey to derive self-consistent and robust membership probabilities of the Pleiades cluster. We aim at developing a methodology to infer membership probabilities to the Pleiades cluster from the DANCe multidimensional astro-photometric data set in a consistent way throughout the entire derivation. The determination of the membership probabilities has to be applicable to censored data and must incorporate the measurement uncertainties into the inference procedure. We use Bayes' theorem and a curvilinear forward model for the likelihood of the measurements of cluster members in the colour-magnitude space, to infer posterior membership probabilities. The distribution of the cluster members proper motions and the distribution of contaminants in the full multidimensional astro-photometric space is modelled with a mixture-of-Gaussians likelihood. We analyse several representation spaces composed of the proper motions plus a subset of the available magnitudes and colour indices. We select two prominent representation spaces composed of variables selected using feature relevance determination techniques based in Random Forests, and analyse the resulting samples of high probability candidates. We consistently find lists of high probability (p > 0.9975) candidates with $\approx$ 1000 sources, 4 to 5 times more than obtained in the most recent astro-photometric studies of the cluster. The methodology presented here is ready for application in data sets that include more dimensions, such as radial and/or rotational velocities, spectral indices and variability.

Adolescent Friendship Networks and Trajectories of Binge Drinking

The authors examine how adolescent friendship networks are linked to drinking trajectories into young adulthood using Add Health. They add to the literature by examining whether an individual's structural position (group member, liaison or isolate) in friendship networks is linked to longitudinal alcohol use, above and beyond number of drinking friends. Trajectories of binge drinking episodes per month are first modeled using semi-parametric longitudinal mixture models. Individuals are assigned to trajectory groups based on posterior probabilities of membership. Friendship network structural characteristics are modeled using NEGOPY. Multinomial logit models of trajectory group membership are then estimated, and include information on network position, number of drinking friends, as well as a range of controls. They identify five trajectories of drinking. Structural position is associated with use trajectories: bingeing group membership and liaison to bingeing groups predicts higher trajectories. Network effects are strongly associated with bingeing in school, but not after.

PWP3D: Real-Time Segmentation and Tracking of 3D Objects

We formulate a probabilistic framework for simultaneous region-based 2D segmentation and 2D to 3D pose tracking, using a known 3D model. Given such a model, we aim to maximise the discrimination between statistical foreground and background appearance models, via direct optimisation of the 3D pose parameters. The foreground region is delineated by the zero-level-set of a signed distance embedding function, and we define an energy over this region and its immediate background surroundings based on pixel-wise posterior membership probabilities (as opposed to likelihoods). We derive the differentials of this energy with respect to the pose parameters of the 3D object, meaning we can conduct a search for the correct pose using standard gradient-based non-linear minimisation techniques. We propose novel enhancements at the pixel level based on temporal consistency and improved online appearance model adaptation. Furthermore, straightforward extensions of our method lead to multi-camera and multi-object tracking as part of the same framework. The parallel nature of much of the processing in our algorithm means it is amenable to GPU acceleration, and we give details of our real-time implementation, which we use to generate experimental results on both real and artificial video sequences, with a number of 3D models. These experiments demonstrate the benefit of using pixel-wise posteriors rather than likelihoods, and showcase the qualities, such as robustness to occlusions and motion blur (and also some failure modes), of our tracker.

Latent class model with familial dependence to address heterogeneity in complex diseases: adapting the approach to family-based association studies

Clinical diagnoses of complex diseases may often encompass multiple genetically heterogeneous disorders. One way of dissecting this heterogeneity is to apply latent class (LC) analysis to measurements related to the diagnosis, such as detailed symptoms, to define more homogeneous disease sub-types, influenced by a smaller number of genes that will thus be more easily detectable. We have previously developed a LC model allowing dependence between the latent disease class status of relatives within families. We have also proposed a strategy to incorporate the posterior probability of class membership of each subject in parametric linkage analysis, which is not directly transferable to genetic association methods. Under the framework of family-based association tests (FBAT), we now propose to make the contribution of an affected subject to the FBAT statistic proportional to his or her posterior class membership probability. Simulations showed a modest but robust power advantage compared to simply assigning each subject to his or her most probable class, and important power gains over the analysis of the disease diagnosis without LC modeling under certain scenarios. The use of LC analysis with FBAT is illustrated using autism spectrum disorder (ASD) symptoms on families from the Autism Genetics Research Exchange, where we examined eight regions previously associated to autism in this sample. The analysis using the posterior probability of membership to an LC detected an association in the JARID2 gene as significant as that for ASD (P = 3 × 10(-5)) but with a larger effect size (odds ratio = 2.17 vs. 1.55).

Adaptive Methods for Spatial Scan Analysis via Semiparametric Mixture Models

Spatial scan density (SSD) estimation via mixture models is an important problem in the field of spatial statistical analysis and has wide applications in image analysis. The “borrowed strength” density estimation (BSDE) method via mixture models enables one to estimate the local probability density function in a random field wherein potential similarities between the density functions for the subregions are exploited. This article proposes an efficient methods for SSD estimation by integrating the borrowed strength technique into the alternative EM framework which combines the statistical basis of the BSDE approach with the stability and improved convergence rate of the alternative EM methods. In addition, we propose adaptive SSD estimation methods that extend the aforementioned approach by eliminating the need to find the posterior probability of membership of the component densities afresh in each subregion. Simulation results and an application to the detection and identification of man-made regions of interest in an unmanned aerial vehicle imagery experiment show that the adaptive methods significantly outperform the BSDE method. Other applications include automatic target recognition, mammographic image analysis, and minefield detection.

Mitochondrial DNA Variation among Wintering Midcontinent Gulf Coast Sandhill Cranes

Based on morphological characteristics, 3 subspecies of sandhill cranes (Grus canadensis)-the lesser sandhill crane (G. c. canadensis), Canadian sandhill crane (G. c. rowani), and greater sandhill crane (G. c. tabida)-occur within the midcontinent population (MCP) and winter along the Gulf Coast of Texas, USA. Of these subspecies, the greater sandhill crane is the least abundant and the most restricted in distribution in the midcontinent region, making it a subspecies of special concern. Twenty individuals from each subspecies with the highest posterior probabilities of membership in subspecies associated with their morphological discriminant score (i.e., those individuals that were least likely to be misclassified based on the morphological model currently used to assign wintering individuals to subspecies) were chosen from a total sample of 220 birds. We amplified and sequenced a 437-base-pair segment from domain I of the mitochondrial DNA (mtDNA) control region for these 60 birds. Analyses of the resulting 41 different mtDNA haplotypes indicate that birds classified as Canadian and greater sandhill cranes based on their morphology do not differ genetically, but lesser sandhill cranes are genetically distinct from both Canadian and greater sandhill cranes. When Canadian and greater sandhill cranes are grouped together, 55 of the 60 individuals sequenced are consistently classified using morphology and mtDNA. We then used the DNA sequences to develop a simple restriction enzyme assay of this mtDNA segment to survey an additional 160 specimens. These results indicate that (1) mtDNA assays provide reliable discrimination of migratory sandhill cranes, (2) the current subspecific designations of sandhill cranes may not accurately reflect genetic structure in this species, (3) concern for preserving the genetic diversity of sandhill cranes in the midcontinent region may not need to focus on management of greater sandhill cranes wintering in the sampled populations, and (4) studies of individuals from the breeding grounds are clearly needed to determine whether Canadian and greater sandhill cranes are indeed distinct subspecies.