When Bayesian latent class analysis is used for diagnostic test data in the absence of a gold standard test, it is common to assume that any unknown test sensitivities and specificities are constant ...across different populations. Indeed this assumption is often necessary for model identifiability. However there are a number of practical situations, depending on the type of test and the nature of the disease, where this assumption may not be true. We present a case study of using a microscopic agglutination test to diagnose leptospiroris infection in beef cattle, which strongly suggests that sensitivity in particular varies among herds. We develop and fit an alternative model in which sensitivity is related to within-herd prevalence, and discuss the statistical and epidemiological implications.
Bayesian mixture models, often termed latent class models, allow users to estimate the diagnostic accuracy of tests and true prevalence in one or more populations when the positive and/or negative ...reference standards are imperfect. Moreover, they allow the data analyst to show the superiority of a novel test over an old test, even if this old test is the (imperfect) reference standard. We use published data on Toxoplasmosis in pigs to explore the effects of numbers of tests, numbers of populations, and dependence structure among tests to ensure model (local) identifiability. We discuss and make recommendations about use of priors, sensitivity analysis, model identifiability and study design options, and strongly argue for the use of Bayesian mixture models as a logical and coherent approach for estimating the diagnostic accuracy of two or more tests.
The Standards for the Reporting of Diagnostic Accuracy (STARD) statement, which was recently updated to the STARD2015 statement, was developed to encourage complete and transparent reporting of test ...accuracy studies. Although STARD principles apply broadly, the checklist is limited to studies designed to evaluate the accuracy of tests when the disease status is determined from a perfect reference procedure or an imperfect one with known measures of test accuracy. However, a reference standard does not always exist, especially in the case of infectious diseases with a long latent period. In such cases, a valid alternative to classical test evaluation involves the use of latent class models that do not require a priori knowledge of disease status. Latent class models have been successfully implemented in a Bayesian framework for over 20 years. The objective of this work was to identify the STARD items that require modification and develop a modified version of STARD for studies that use Bayesian latent class analysis to estimate diagnostic test accuracy in the absence of a reference standard. Examples and elaborations for each of the modified items are provided. The new guidelines, termed STARD-BLCM (Standards for Reporting of Diagnostic accuracy studies that use Bayesian Latent Class Models), will facilitate improved quality of reporting on the design, conduct and results of diagnostic accuracy studies that use Bayesian latent class models.
We discuss the issue of identifiability of models for multiple dichotomous diagnostic tests in the absence of a gold standard (GS) test. Data arise as multinomial or product-multinomial counts ...depending upon the number of populations sampled. Models are generally posited in terms of population prevalences, test sensitivities and specificities, and test dependence terms. It is commonly believed that if the degrees of freedom in the data meet or exceed the number of parameters in a fitted model then the model is identifiable. Goodman (1974, Biometrika 61, 215-231) established that this was not the case a long time ago. We discuss currently available models for multiple tests and argue in favor of an extension of a model that was developed by Dendukuri and Joseph (2001, Biometrics 57, 158-167). Subsequently, we further develop Goodman's technique, and make geometric arguments to give further insight into the nature of models that lack identifiability. We present illustrations using simulated and real data.
Wagenmakers, Wetzels, Borsboom, and van der Maas (2011)
argued that psychologists should replace the familiar "frequentist" statistical analyses of their data with Bayesian analyses. To illustrate ...their argument, they reanalyzed a set of psi experiments published recently in this journal by
Bem (2011)
, maintaining that, contrary to his conclusion, his data do not yield evidence in favor of the psi hypothesis. We argue that they have incorrectly selected an unrealistic prior distribution for their analysis and that a Bayesian analysis using a more reasonable distribution yields strong evidence in favor of the psi hypothesis. More generally, we argue that there are advantages to Bayesian analyses that merit their increased use in the future. However, as Wagenmakers et al.'s analysis inadvertently revealed, they contain hidden traps that must be better understood before being more widely substituted for the familiar frequentist analyses currently employed by most research psychologists.
Essentially all women are exposed to polycyclic aromatic hydrocarbons (PAHs), formed during incomplete combustion of organic materials, including fossil fuels, wood, foods, and tobacco. PAHs are ...ovarian toxicants in rodents, and cigarette smoking is associated with reproductive abnormalities in women. Biomonitoring of hydroxylated PAH (OH-PAH) metabolites in urine provides an integrated measure of exposure to PAHs via multiple routes and has been used to characterize exposure to PAHs in humans. We hypothesized that concentrations of OH-PAHs in urine are associated with reproductive function in women. We recruited women 18–44years old, living in Orange County, California to conduct daily measurement of urinary luteinizing hormone (LH) and estrone 3-glucuronide (E13G) using a microelectronic fertility monitor for multiple menstrual cycles; these data were used to calculate endocrine endpoints. Participants also collected urine samples on cycle day 10 for measurement of nine OH-PAHs. Models were constructed for eight endpoints using a Bayesian mixed modeling approach with subject-specific random effects allowing each participant to act as a baseline for her set of measurements. We observed associations between individual OH-PAH concentrations and follicular phase length, follicular phase LH and E13G concentrations, preovulatory LH surge concentrations, and periovulatory E13G slope and concentration. We have demonstrated the feasibility of using urinary reproductive hormone data obtained via fertility monitors to calculate endocrine endpoints for epidemiological studies of ovarian function during multiple menstrual cycles. The results show that environmental exposure to PAHs is associated with changes in endocrine markers of ovarian function in women in a PAH-specific manner.
•Urinary reproductive hormones were measured using fertility monitors in cycling women.•Urinary metabolites of polycyclic aromatic hydrocarbons were measured once per cycle.•Environmental exposure to PAHs is associated with follicular phase length, follicular LH and E13G concentrations, preovulatory LH levels, and periovulatory E13G.
Split Hamiltonian Monte Carlo Shahbaba, Babak; Lan, Shiwei; Johnson, Wesley O. ...
Statistics and computing,
05/2014, Letnik:
24, Številka:
3
Journal Article
Recenzirano
Odprti dostop
We show how the Hamiltonian Monte Carlo algorithm can sometimes be speeded up by “splitting” the Hamiltonian in a way that allows much of the movement around the state space to be done at low ...computational cost. One context where this is possible is when the log density of the distribution of interest (the potential energy function) can be written as the log of a Gaussian density, which is a quadratic function, plus a slowly-varying function. Hamiltonian dynamics for quadratic energy functions can be analytically solved. With the splitting technique, only the slowly-varying part of the energy needs to be handled numerically, and this can be done with a larger stepsize (and hence fewer steps) than would be necessary with a direct simulation of the dynamics. Another context where splitting helps is when the most important terms of the potential energy function and its gradient can be evaluated quickly, with only a slowly-varying part requiring costly computations. With splitting, the quick portion can be handled with a small stepsize, while the costly portion uses a larger stepsize. We show that both of these splitting approaches can reduce the computational cost of sampling from the posterior distribution for a logistic regression model, using either a Gaussian approximation centered on the posterior mode, or a Hamiltonian split into a term that depends on only a small number of critical cases, and another term that involves the larger number of cases whose influence on the posterior distribution is small.
We develop a dependent Dirichlet process model for survival analysis data. A major feature of the proposed approach is that there is no necessity for resulting survival curve estimates to satisfy the ...ubiquitous proportional hazards assumption. An illustration based on a cancer clinical trial is given, where survival probabilities for times early in the study are estimated to be lower for those on a high-dose treatment regimen than for those on the low dose treatment, while the reverse is true for later times, possibly due to the toxic effect of the high dose for those who are not as healthy at the beginning of the study.
We explored factors associated with reasons that women with urinary incontinence (UI) reported for not seeking treatment for their UI from a healthcare professional and whether reasons differed by ...race/ethnicity, socioeconomic status, or education.
We analyzed questionnaire data collected from 1995 to 2005 in the Study of Women's Health Across the Nation. In visits 7 to 9, we elicited reasons that women with UI reported for not seeking treatment and condensed them into: UI not bad enough, beliefs about UI causes (UI is a normal consequence of aging or childbirth), and motivational barriers (such as feeling too embarrassed). We used Generalized Estimating Equations and ordinal logistic regression to evaluate factors associated with these reported reasons and number of reasons.
Of the 1,339 women reporting UI, 814 (61.0%) reported they did not seek treatment for UI. The most frequently reported reasons were as follows: "UI not bad enough" (73%), "UI is a normal part of aging" (53%), and "healthcare provider never asked" (55%). Women reporting daily UI had higher odds of reporting beliefs about UI causes (adjusted odds ratio UI 3.16, 95% CI 1.64-6.11) or motivational barriers (adjusted odds ratio UI 2.36, 95% CI 1.21-4.63) compared with women reporting less than monthly UI. We found no interactions by race/ethnicity, socioeconomic status, or education and UI characteristics in reasons that women reported for not seeking UI treatment.
Over half of women who did not seek treatment for their UI reported reasons that could be addressed by public health and clinical efforts to make UI a discussion point during midlife well-women visits.