RESUMO
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.
Assuntos
Algoritmos , Teorema de Bayes , Modelos Estatísticos , Humanos , Feminino , Pessoa de Meia-Idade , Idoso , Simulação por Computador , Método de Monte Carlo , Funções Verossimilhança , Cadeias de MarkovRESUMO
BACKGROUND: Identifying children at high risk of developing asthma can facilitate prevention and early management strategies. We developed a prediction model of children's asthma risk using objectively collected population-based children and parental histories of comorbidities. METHODS: We conducted a retrospective population-based cohort study using administrative data from Manitoba, Canada, and included children born from 1974 to 2000 with linkages to ≥1 parent. We identified asthma and prior comorbid condition diagnoses from hospital and outpatient records. We used two machine-learning models: least absolute shrinkage and selection operator (LASSO) logistic regression (LR) and random forest (RF) to identify important predictors. The predictors in the base model included children's demographics, allergic conditions, respiratory infections, and parental asthma. Subsequent models included additional multiple comorbidities for children and parents. RESULTS: The cohort included 195,666 children: 51.3% were males and 17.7% had asthma diagnosis. The base LR model achieved a low predictive performance with sensitivity of 0.47, 95% confidence interval (0.45-0.48), and specificity of 0.67 (0.66-0.67) using a predicted probability threshold of 0.20. Sensitivity significantly improved when children's comorbidities were included using LASSO LR: 0.71 (0.69-0.72). Predictive performance further improved by including parental comorbidities (sensitivity = 0.72 [0.70-0.73], specificity = 0.69 [0.69-0.70]). We observed similar results for the RF models. Children's menstrual disorders and mood and anxiety disorders, parental lipid metabolism disorders and asthma were among the most important variables that predicted asthma risk. CONCLUSION: Including children and parental comorbidities to children's asthma prediction models improves their accuracy.
Assuntos
Asma , Masculino , Feminino , Humanos , Criança , Estudos de Coortes , Estudos Retrospectivos , Asma/diagnóstico , Asma/epidemiologia , Transtornos de Ansiedade , CanadáRESUMO
PURPOSE: Item non-response (i.e., missing data) may mask the detection of differential item functioning (DIF) in patient-reported outcome measures or result in biased DIF estimates. Non-response can be challenging to address in ordinal data. We investigated an unsupervised machine-learning method for ordinal item-level imputation and compared it with commonly-used item non-response methods when testing for DIF. METHODS: Computer simulation and real-world data were used to assess several item non-response methods using the item response theory likelihood ratio test for DIF. The methods included: (a) list-wise deletion (LD), (b) half-mean imputation (HMI), (c) full information maximum likelihood (FIML), and (d) non-negative matrix factorization (NNMF), which adopts a machine-learning approach to impute missing values. Control of Type I error rates were evaluated using a liberal robustness criterion for α = 0.05 (i.e., 0.025-0.075). Statistical power was assessed with and without adoption of an item non-response method; differences > 10% were considered substantial. RESULTS: Type I error rates for detecting DIF using LD, FIML and NNMF methods were controlled within the bounds of the robustness criterion for > 95% of simulation conditions, although the NNMF occasionally resulted in inflated rates. The HMI method always resulted in inflated error rates with 50% missing data. Differences in power to detect moderate DIF effects for LD, FIML and NNMF methods were substantial with 50% missing data and otherwise insubstantial. CONCLUSION: The NNMF method demonstrated comparable performance to commonly-used non-response methods. This computationally-efficient method represents a promising approach to address item-level non-response when testing for DIF.
Assuntos
Medidas de Resultados Relatados pelo Paciente , Qualidade de Vida , Simulação por Computador , Humanos , Funções Verossimilhança , Psicometria/métodos , Qualidade de Vida/psicologiaRESUMO
BACKGROUND: Although memory and cognitive declines are associated with normal brain aging, they may also be precursors to dementia. OBJECTIVE: We aim to offer a novel approach to prevent or slow the progress of neurodegenerative dementia, or plausibly, improve the cognitive functions of individuals with dementia. METHODS: We will recruit and enroll 75 participants (older than 50 years old with either mild cognitive impairment or probable early or moderate dementia) for this double-blind randomized controlled study to estimate the efficacy of active transcranial alternating current stimulation with cognitive treatment (in comparison with sham transcranial alternating current stimulation). This will be a crossover study; a cycle consists of sham or active treatment for a period of 4 weeks (5 days per week, in two 30-minute sessions with a half-hour break in between), and participants are randomized into 2 groups, with stratification by age, sex, and cognitive level (measured with the Montreal Cognitive Assessment). Outcomes will be assessed before and after each treatment cycle. The primary outcomes are changes in Wechsler Memory Scale Older Adult Battery and Alzheimer Disease Assessment Scale scores. Secondary outcomes are changes in performance on tests of frontal lobe functioning (verbal fluency), neuropsychiatric symptoms (Neuropsychiatric Inventory Questionnaire), mood changes (Montgomery-Åsberg Depression Rating Scale), and short-term recall (visual 1-back task). Exploratory outcome measures will also be assessed: static and dynamic vestibular response using electrovestibulography, neuronal changes using functional near-infrared spectroscopy, and change in spatial orientation using virtual reality navigation. RESULTS: As of February 10, 2022, the study is ongoing: 7 patients have been screened, and all were deemed eligible for and enrolled in the study; 4 participants have completed baseline assessments. CONCLUSIONS: We anticipate that transcranial alternating current stimulation will be a well-tolerated treatment, with no serious side effects and with considerable short- and long-term cognitive improvements. TRIAL REGISTRATION: Clinicaltrials.gov NCT05203523; https://clinicaltrials.gov/show/NCT05203523. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/37282.
RESUMO
The majority of the research on rank-based sampling designs in finite populations has been concerned with univariate situations. In this article, we study design-based estimation using a bivariate ranked set sampling (BIRSS) for finite populations when we have bivariate response variables. We derive the first and second-order inclusion probabilities associated with a BIRSS design. We show that the size of a BIRSS sample is random and propose using a conditional Poisson sampling (CPS) design to rectify this problem. We then use calculated inclusion probabilities to obtain design-based estimators of correlation coefficients between the bone mineral density (BMD) levels at the baseline and followup of a longitudinal BMD study in the province of Manitoba in Canada. We also study the problem of estimating the parameters of a regression model between the followup BMD and easy to obtain auxiliary information from the underlying population. Finally, we study the problem of classifying patients as those with or without osteoporosis using BIRSS and various CPS designs. We show that BIRSS designs are very flexible and can be used to obtain more efficient design-based estimators in sample surveys when dealing with response variables that are hard to measure or expensive to obtain.
Assuntos
Osteoporose , Projetos de Pesquisa , Densidade Óssea , Humanos , Manitoba , ProbabilidadeRESUMO
Statistical learning with ranked set samples has shown promising results in estimating various population parameters. Despite the vast literature on rank-based statistical learning methodologies, very little effort has been devoted to studying regression analysis with such samples. A pressing issue is how to incorporate the rank information of ranked set samples into the analysis. We propose two methodologies based on a weighted least squares approach and multilevel modeling to better incorporate the rank information of such samples into the estimation and prediction processes of regression-type models. Our approaches reveal significant improvements in both estimation and prediction problems over already existing methods in the literature and the corresponding ones with simple random samples. We study the robustness of our methods with respect to the misspecification of the distribution of the error terms. Also, we show that rank-based regression models can effectively predict simple random test data by assigning ranks to them a posteriori using judgment poststratification. Theoretical results are augmented with simulations and an osteoporosis study based on a real data set from the Bone Mineral Density (BMD) program of Manitoba to estimate the BMD level of patients using easy to obtain covariates.
Assuntos
Osteoporose , Humanos , Densidade Óssea , Análise de RegressãoRESUMO
BACKGROUND: Many clinical trials investigating treatment efficacy require an interim analysis. Recently we have been running a large, multisite, randomized, placebo-controlled, double-blind clinical trial investigating the effect of repetitive transcranial magnetic stimulation (rTMS) treatment for improving or stabilizing the cognition of patients diagnosed with Alzheimer disease. OBJECTIVE: The objectives of this paper are to report on recruitment, adherence, and adverse events (AEs) to date, and to describe in detail the protocol for interim analysis of the clinical trial data. The protocol will investigate whether the trial is likely to reach its objectives if continued to the planned maximum sample size. METHODS: The specific requirements of the analytic protocol are to (1) ensure the double-blind nature of the data while doing the analysis, (2) estimate the predictive probabilities of success (PPoSs), (3) estimate the numbers needed to treat, (4) re-estimate the initial required sample size. The initial estimate of sample size was 208. The interim analysis will be based on 150 patients who will be enrolled in the study and finish at least 8 weeks of the study. Our protocol for interim analysis, at the very first stage, is to determine the response rate for each participant to the treatment (either sham or active), while ensuring the double-blind nature of the data. The blinded data will be analyzed by a statistician to investigate the treatment efficacy. We will use Bayesian PPoS to predict the success rate and determine whether the study should continue. RESULTS: The enrollment has been slowed significantly due to the COVID-19 pandemic and lockdown. Nevertheless, so far 133 participants have been enrolled, while 22 of these have been withdrawn or dropped out for various reasons. In general, rTMS has been found tolerable with no serious AE. Only 2 patients dropped out of the study due to their intolerability to rTMS pulses. CONCLUSIONS: Overall, the study with the same protocol is going as expected with no serious AE or any major protocol deviation. TRIAL REGISTRATION: ClinicalTrials.gov NCT02908815; https://clinicaltrials.gov/ct2/show/NCT02908815. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): DERR1-10.2196/31183.
RESUMO
Quantifying the tool-tissue interaction forces in surgery can be used in the training process of novice surgeons to help them better handle surgical tools and avoid exerting excessive forces. A significant challenge concerns the development of proper statistical learning techniques to model the relationship between the true force exerted on the tissue and several outputs read from sensors mounted on the surgical tools. We propose a nonparametric bootstrap technique and a Bayesian multilevel modeling methodology to estimate the true forces. We use the linear exponential loss function to asymmetrically penalize the over and underestimation of the applied forces to the tissue. We incorporate the direction of the force as a group factor in our analysis. A weighted approach is used to account for the nonhomogeneity of read voltages from the surgical tool. Our proposed Bayesian multilevel models provide estimates that are more accurate than those under the maximum likelihood and restricted maximum likelihood approaches. Moreover, confidence bounds are much narrower and the biases and root mean squared errors are significantly smaller in our multilevel models with the linear exponential loss function.
Assuntos
Calibragem , Teorema de Bayes , Funções VerossimilhançaRESUMO
BACKGROUND: Abdominal aortic calcification (AAC) identified on dual-energy x-ray absorptiometry (DXA) vertebral fracture assessment (VFA) lateral spine images is predictive of cardiovascular outcomes, but is time-consuming to perform manually. Whether this procedure can be automated using convolutional neural networks (CNNs), a class of machine learning algorithms used for image processing, has not been widely investigated. METHODS: Using the Province of Manitoba Bone Density Program DXA database, we selected a random sample of 1100 VFA images from individuals qualifying for VFA as part of their osteoporosis assessment. For each scan, AAC was manually scored using the 24-point semi-quantitative scale and categorized as low (score < 2), moderate (score 2 to <6), or high (score ≥ 6). An ensemble consisting of two CNNs was developed, by training and evaluating separately on single-energy and dual-energy images. AAC prediction was performed using the mean AAC score of the two models. RESULTS: Mean (SD) age of the cohort was 75.5 (6.7) years, 95.5% were female. Training (N = 770, 70%), validation (N = 110, 10%) and test sets (N = 220, 20%) were well-balanced with respect to baseline characteristics and AAC scores. For the test set, the Pearson correlation between the CNN-predicted and human-labelled scores was 0.93 with intraclass correlation coefficient for absolute agreement 0.91 (95% CI 0.89-0.93). Kappa for AAC category agreement (prevalence- and bias-adjusted, ordinal scale) was 0.71 (95% CI 0.65-0.78). There was complete separation of the low and high categories, without any low AAC score scans predicted to be high and vice versa. CONCLUSIONS: CNNs are capable of detecting AAC in VFA images, with high correlation between the human and predicted scores. These preliminary results suggest CNNs are a promising method for automatically detecting and quantifying AAC.
Assuntos
Fraturas da Coluna Vertebral , Calcificação Vascular , Absorciometria de Fóton , Idoso , Aorta Abdominal/diagnóstico por imagem , Densidade Óssea , Feminino , Humanos , Aprendizado de Máquina , Manitoba , Projetos Piloto , Calcificação Vascular/diagnóstico por imagemRESUMO
Judgment post-stratification is used to supplement observations taken from finite mixture models with additional easy to obtain rank information and incorporate it in the estimation of model parameters. To do this, sampled units are post-stratified on ranks by randomly selecting comparison sets for each unit from the underlying population and assigning ranks to them using available auxiliary information or judgment ranking. This results in a set of independent order statistics from the underlying model, where the number of units in each rank class is random. We consider cases where one or more rankers with different ranking abilities are used to provide judgment ranks. The judgment ranks are then combined to produce a strength of agreement measure for each observation. This strength measure is implemented in the maximum likelihood estimation of model parameters via a suitable expectation maximization algorithm. Simulation studies are conducted to evaluate the performance of the estimators with or without the extra rank information. Results are applied to bone mineral density data from the third National Health and Nutrition Examination Survey to estimate the prevalence of osteoporosis in adult women aged 50 and over.
Assuntos
Modelos Estatísticos , Osteoporose Pós-Menopausa/epidemiologia , Absorciometria de Fóton , Algoritmos , Biomarcadores , Densidade Óssea , Feminino , Humanos , Julgamento , Funções Verossimilhança , Pessoa de Meia-Idade , Osteoporose Pós-Menopausa/diagnóstico por imagem , PrevalênciaRESUMO
This paper studies quantile regression analysis with maxima or minima nomination sampling designs. These designs are often used to obtain more representative samples from the tails of the underlying distribution using the easy to access rank information during the sampling process. We propose new loss functions to incorporate the rank information of nominated samples in the estimation process. Also, we provide an alternative approach that translates estimation problems with nominated samples to corresponding problems under simple random sampling (SRS). Strategies are given to choose proper nomination sampling designs for a given population quantile. Numerical studies show that quantile regression models with maxima (or minima) nominated samples have higher relative efficiencies compared with their counterparts under SRS for analyzing the upper (or lower) tail quantiles of the distribution of the response variable. Results are then implemented on a large cohort study in the Canadian province of Manitoba to analyze quantiles of bone mineral density using available covariates. We show that in some cases, methods based on nomination sampling designs require about one-tenth of the sample used in SRS to estimate the lower or upper tail conditional quantiles with comparable mean squared errors. This is a dramatic reduction in time and cost compared with the usual SRS approach.
Assuntos
Análise de Regressão , Tamanho da Amostra , Idoso , Densidade Óssea , Simulação por Computador , Análise Custo-Benefício , Humanos , Funções Verossimilhança , Masculino , Manitoba , Pessoa de Meia-Idade , TempoRESUMO
Knowledge of forces, exerted on the brain tissue during the performance of neurosurgical tasks, is critical for quality assurance, case rehearsal, and training purposes. Quantifying the interaction forces has been made possible by developing SmartForceps, a bipolar forceps retrofitted by a set of strain gauges. The forces are estimated using voltages read from strain gauges. We therefore need to quantify the force-voltage relationship to estimate the interaction forces during microsurgery. This problem has been addressed in the literature by following the physical and deterministic properties of the force-sensing strain gauges without obtaining the precision associated with each estimate. In this paper, we employ a probabilistic methodology by using a nonparametric Bootstrap approach to obtain both point and interval estimates of the applied forces at the tool tips, while the precision associated with each estimate is provided. To show proof-of-concept, the Bootstrap technique is employed to estimate unknown forces, and construct necessary confidence intervals using observed voltages in data sets that are measured from the performance of surgical tasks on a cadaveric brain. Results indicate that the Bootstrap technique is capable of estimating tool-tissue interaction forces with acceptable level of accuracy compared to the linear regression technique under the normality assumption.
Assuntos
Encéfalo/cirurgia , Microcirurgia/instrumentação , Procedimentos Neurocirúrgicos/instrumentação , Instrumentos Cirúrgicos , Calibragem , Humanos , Análise dos Mínimos Quadrados , Microcirurgia/métodos , Procedimentos Neurocirúrgicos/métodos , Pressão , Estatísticas não ParamétricasRESUMO
Rank-based sampling designs are widely used in situations where measuring the variable of interest is costly but a small number of sampling units (set) can be easily ranked prior to taking the final measurements on them and this can be done at little cost. When the variable of interest is binary, a common approach for ranking the sampling units is to estimate the probabilities of success through a logistic regression model. However, this requires training samples for model fitting. Also, in this approach once a sampling unit has been measured, the extra rank information obtained in the ranking process is not used further in the estimation process. To address these issues, in this paper, we propose to use the partially rank-ordered set sampling design with multiple concomitants. In this approach, instead of fitting a logistic regression model, a soft ranking technique is employed to obtain a vector of weights for each measured unit that represents the probability or the degree of belief associated with its rank among a small set of sampling units. We construct an estimator which combines the rank information and the observed partially rank-ordered set measurements themselves. The proposed methodology is applied to a breast cancer study to estimate the proportion of patients with malignant (cancerous) breast tumours in a given population. Through extensive numerical studies, the performance of the estimator is evaluated under various concomitants with different ranking potentials (i.e. good, intermediate and bad) and tie structures among the ranks. We show that the precision of the partially rank-ordered set estimator is better than its counterparts under simple random sampling and ranked set sampling designs and, hence, the sample size required to achieve a desired precision is reduced.