Violations Of Model Assumptions Research Articles

Rosace: a robust deep mutational scanning analysis framework employing position and mean-variance shrinkage

Deep mutational scanning (DMS) measures the effects of thousands of genetic variants in a protein simultaneously. The small sample size renders classical statistical methods ineffective. For example, p-values cannot be correctly calibrated when treating variants independently. We propose Rosace, a Bayesian framework for analyzing growth-based DMS data. Rosace leverages amino acid position information to increase power and control the false discovery rate by sharing information across parameters via shrinkage. We also developed Rosette for simulating the distributional properties of DMS. We show that Rosace is robust to the violation of model assumptions and is more powerful than existing tools.

Causal relationship between dietary salt intake and dementia risk: Mendelian randomization study.

Observational research has indicated a potential link between dietary salt intake and susceptibility to dementia. However, it is important to note that these types of studies are prone to the issues of reverse causation and residual confounding. Therefore, we conducted a two-sample Mendelian randomization (MR) study to explore the causality. To explore the causal relationship between them, this Mendelian randomization (MR) study incorporated summary statistics of dietary salt intake and dementia. We estimated the causality between salt intake and the risk of overall dementia andvarious subtypes of dementia, including Alzheimer's disease (AD), Vascular dementia (VaD), and Lewy body dementia (LBD). The inverse variance-weighted (IVW) method was the major MR analysis. To conduct sensitivity analyses, we employed various MR methods, the pleiotropy residual sum and outlier (MR-PRESSO) method, and the leave-one-out approach. The MR-Egger intercept and Cochran's Q test were conducted to test pleiotropy and heterogeneityrespectively. A suggestive association was observed for genetically predicted higher dietary salt intake and increased risk of overalldementia in the European ancestry [odds ratio (OR): 1.542; 95% confidence interval (95% CI): 1.095-2.169; P = 0.013]. The causal relationship between dietary salt intake and overalldementia is robust with respect to the choice of statistical methods and is validated through extensive sensitivity analyses that guard against various model assumption violations. Meanwhile, no clear heterogeneity or pleiotropy was identified. However, we failed to detect a causal effect of dietary salt intake on the risk of various dementia subtypes. The results of this research present strong evidence that established a significant association between dietary salt intake and the likelihood of developing dementia. These findings reinforce the notion that the amount of dietary salt intake plays a crucial role in determining the risk of acquiring this cognitive condition. By establishing a definitive correlation, this study highlights the importance of reducing salt consumption as a preventive measure against dementia.

Efficacy of propofol-supplemented cardioplegia on biomarkers of organ injury in patients having cardiac surgery using cardiopulmonary bypass: a statistical analysis plan for the ProMPT-2 randomised controlled trial

BackgroundThe ProMPT-2 trial (Propofol for Myocardial Protection Trial #2) aims to compare the safety and efficacy of low- and high-dose propofol supplementation of the cardioplegia solution during adult cardiac surgery versus sham supplementation. This update presents the statistical analysis plan, detailing how the trial data will be analysed and presented. Outlined analyses are in line with the Consolidated Standards of Reporting Trials and the statistical analysis plan has been written prior to database lock and the final analysis of trial data to avoid reporting bias (following recommendations from the International Conference on Harmonisation of Good Clinical Practice).Methods/designProMPT-2 is a multi-centre, blinded, parallel three-group randomised controlled trial aiming to recruit 240 participants from UK cardiac surgery centres to either sham cardioplegia supplementation, low dose (6 µg/ml) or high dose (12 µg/ml) propofol cardioplegia supplementation. The primary outcome is cardiac-specific troponin T levels (a biomarker of cardiac injury) measured during the first 48 h following surgery.The statistical analysis plan describes the planned analyses of the trial primary and secondary outcomes in detail, including approaches to deal with missing data, multiple testing, violation of model assumptions, withdrawals from the trial, non-adherence with the treatment and other protocol deviations. It also outlines the planned sensitivity analyses and exploratory analyses to be performed.DiscussionThis manuscript prospectively describes, prior to the completion of data collection and database lock, the analyses to be undertaken for the ProMPT-2 trial to reduce risk of reporting and data-driven analyses.Trial registrationISRCTN ISRCTN15255199. Registered on 26 March 2019.

Genetic causal association between varicella-zoster virus infection and psychiatric disorders: A 2-sample Mendelian randomization study

BackgroundPsychiatric disorders, such as schizophrenia (SCZ), major depressive (MDD), and bipolar disorder (BD) have a profound impact on millions of individuals worldwide. The critical step toward developing effective preventive and treatment strategies lies in comprehending the causal mechanisms behind these diseases and identifying modifiable risk factors associated with them. MethodsIn this study, we conducted a 2-sample Mendelian randomization analysis to explore the potential links between chickenpox(varicella-zoster virus infection) and three major psychiatric disorders(SCZ, MDD, BD). ResultsIn our MR study, among the three major psychiatric disorders, chickenpox was shown to be causally related to BD, indicating that infection with chickenpox may increase the risk of developing BD (IVW: OR = 1.064, 95% CI =1.025–1.104, P=0.001; RAPS: OR=1.066, 95% CI=1.024–1.110, P=0.002), while there was no causal relationship between SCZ and MDD. Similar estimated causal effects were observed consistently across the various MR models. The robustness of the identified causal relationship between chickenpox and BD holds true regardless of the statistical methods employed, as confirmed by extensive sensitivity analyses that address violations in model assumptions. The MR–Egger regression test failed to reveal any signs of directional pleiotropy (intercept = −0.042, standard error (SE) = 0.029, p = 0.236). Similarly, the MR-PRESSO analysis revealed no evidence of directional pleiotropy or outliers among the chickenpox-related instrumental variables (global test p = 0.653). Furthermore, a leave-one-out sensitivity analysis yielded consistent results, further underscoring the credibility and stability of the causal relationship. ConclusionsOur findings provide compelling evidence of a causal effect of chickenpox on the risk of BD. To gain a more comprehensive understanding of this association and its underlying mechanisms, additional research is needed. Such investigations are pivotal in identifying effective interventions for promoting BD prevention.

Generating normative data from web-based administration of the Cambridge Neuropsychological Test Automated Battery using a Bayesian framework.

Normative cognitive data can distinguish impairment from healthy cognitive function and pathological decline from normal ageing. Traditional methods for deriving normative data typically require extremely large samples of healthy participants, stratifying test variation by pre-specified age groups and key demographic features (age, sex, education). Linear regression approaches can provide normative data from more sparsely sampled datasets, but non-normal distributions of many cognitive test results may lead to violation of model assumptions, limiting generalisability. The current study proposes a novel Bayesian framework for normative data generation. Participants (n = 728; 368 male and 360 female, age 18-75 years), completed the Cambridge Neuropsychological Test Automated Battery via the research crowdsourcing website Prolific.ac. Participants completed tests of visuospatial recognition memory (Spatial Working Memory test), visual episodic memory (Paired Associate Learning test) and sustained attention (Rapid Visual Information Processing test). Test outcomes were modelled as a function of age using Bayesian Generalised Linear Models, which were able to derive posterior distributions of the authentic data, drawing from a wide family of distributions. Markov Chain Monte Carlo algorithms generated a large synthetic dataset from posterior distributions for each outcome measure, capturing normative distributions of cognition as a function of age, sex and education. Comparison with stratified and linear regression methods showed converging results, with the Bayesian approach producing similar age, sex and education trends in the data, and similar categorisation of individual performance levels. This study documents a novel, reproducible and robust method for describing normative cognitive performance with ageing using a large dataset.

Leveraging Angler Effort to Inform Fisheries Management: Using Harvest and Harvest Rate to Estimate Abundance of White Sturgeon

Abstract Traditional methods for estimating abundance of fish populations are not feasible in some systems due to complex population structure and constraints on sampling effort. Lincoln’s estimator provides a technique that uses harvest and harvest rate to estimate abundance. Using angler catch data allows assumptions of the estimator to be addressed without relying on methods that could be prohibitively field-intensive or costly. Historic estimates of White Sturgeon Acipenser transmontanus abundance in the Sacramento–San Joaquin River basin have been obtained using mark–recapture methods; however, White Sturgeon population characteristics often cause violations of model assumptions, such as population closure and independent capture probabilities. We developed a version of Lincoln’s estimator using a joint likelihood, estimated abundance of White Sturgeon in the Sacramento–San Joaquin River basin in 2015 using this method and empirical data and assessed accuracy and precision of estimates in a simulation study. Estimating abundance using harvest and harvest rate, as represented by our model framework, has the potential to be precise and accurate. The joint likelihood–based approach fitted using Bayesian methods is advantageous because it includes all sources of variation in a single model. Precision of abundance estimates was low with application of the model to White Sturgeon in the Sacramento–San Joaquin River basin and to similar conditions in a simulated dataset. Using simulation, precision and accuracy increased with increases in the number of high-reward and standard tags released, tag reporting rate, tag retention rate, and harvest rate. Results demonstrate potential sources of error when using this approach and suggest that increasing the number of tagged fish and tag reporting rate are potential actions to improve precision and accuracy of abundance estimates of the model.

A demonstration of the value of recapture data for informing moult phenology models for avian species with imperfect moult data

The Underhill–Zucchini model has revolutionised the study of moult in the context of birds’ annual cycles, but, as for all statistical models, inferences are vulnerable to violations of model assumptions. In particular, the standard Underhill–Zucchini models for moult phenology are vulnerable to imperfect moult data arising, for example, from non-random sampling and/or the misclassification of non-moulting birds. Similarly, inference is challenging for species with dispersed moult periods (population-level moult extending beyond an annual cycle). Using ringing data from Cape Weavers Ploceus capensis and SAFRING data for Cape Sugarbirds Promerops cafer, we demonstrate how recent extensions to the Underhill–Zucchini framework can help the robust estimation of moult parameters in such situations, in particular when within-season recapture data is available.

Density estimates of unmarked mammals: comparing two models and assumptions across multiple species and years

Density estimation is a key goal in ecology, but accurate estimates for unmarked animals remain elusive. Camera trap data can bridge this gap, but accuracy, precision, and concordance varies among estimators. We compared estimates from unmarked spatial capture–recapture (spatial count (SC)) models, and time in front of camera (TIFC) models, for four large mammal species in boreal Canada. Species differed in movement rates, behaviours, and sociality—traits related to model assumptions. TIFC densities typically exceeded SC model estimates for all species. Two- to five-fold differences between estimators were common. SC estimates were annually stable for moose and caribou but not for white-tailed deer. TIFC estimates showed high annual variation in some species, sites, and years, and consistency in others. Both models often produced imprecise estimates. Estimates varied from DNA- and aerial survey-based estimates. We contend models diverge, or implausibly vary, due to violations of model assumptions incurred by animal behaviour. Gregarious animals pose challenges to SC, whereas curious animals pose challenges for TIFC models. Simulations can help unravel the role of assumption violations in affecting accuracy of estimates, but field applications across species and landscapes help interpret the outcomes of estimating density from simulated data.

Regulatory Guidance on Randomization and the Use of Randomization Tests in Clinical Trials: A Systematic Review

Randomization is a critical aspect of the clinical trial. Regulatory guidance plays a role in clinical research. Randomization tests provide valid methodology for statistical inference. In the present paper, we intended to survey: (a) information contained in regulatory guidance on randomization; and (b) the use of randomization tests in trials supporting marketing applications. This systematic review used the Cortellis Regulatory Intelligence database (IDRAC). For (a), of n = 156 guidance documents, nine provided recommendations on randomization methods. For (b), n = 48 trials (52 submissions) employed randomization tests. 40 (83.3%) were phase III trials, 31 (64.6%) employed dynamic allocation. Randomization test was the primary in 5 (10.4%), and sensitivity analysis in 34 (70.8%) trials. Furthermore, randomization tests were performed when treatment allocation was not according to plan (4 (8.3%)), or violations of model assumptions (5 (10.4%)) were observed. Overall, there are sparse recommendations on randomization in current regulatory guidance, with little mention of recently developed methods which might outperform current practice in certain contexts. Greater adoption of such methods in pharmaceutical industry trials may lead to a better shared sense between regulators and trial sponsors on their appropriate use. Randomization tests are employed with greater frequency, reflecting increased appreciation of the validity and value of randomization tests.

CNETML: maximum likelihood inference of phylogeny from copy number profiles of multiple samples

Phylogenetic trees based on copy number profiles from multiple samples of a patient are helpful to understand cancer evolution. Here, we develop a new maximum likelihood method, CNETML, to infer phylogenies from such data. CNETML is the first program to jointly infer the tree topology, node ages, and mutation rates from total copy numbers of longitudinal samples. Our extensive simulations suggest CNETML performs well on copy numbers relative to ploidy and under slight violation of model assumptions. The application of CNETML to real data generates results consistent with previous discoveries and provides novel early copy number events for further investigation.

Influential Observations in Bayesian Regression Tree Models

Bayesian Classification and Regression Trees (BCART) and Bayesian Additive Regression Trees (BART) are popular Bayesian regression models widely applicable in modern regression problems. Their popularity is intimately tied to the ability to flexibly model complex responses depending on high-dimensional inputs while simultaneously being able to quantify uncertainties. This ability to quantify uncertainties is key, as it allows researchers to perform appropriate inferential analyses in settings that have generally been too difficult to handle using the Bayesian approach. However, surprisingly little work has been done to evaluate the sensitivity of these modern regression models to violations of modeling assumptions. In particular, we will consider influential observations, which one reasonably would imagine to be common—or at least a concern—in the big-data setting. In this article, we consider both the problem of detecting influential observations and adjusting predictions to not be unduly affected by such potentially problematic data. We consider three detection diagnostics for Bayesian tree models, one an analogue of Cook’s distance and the others taking the form of a divergence measure and a conditional predictive density metric, and then propose an importance sampling algorithm to re-weight previously sampled posterior draws so as to remove the effects of influential data in a computationally efficient manner. Finally, our methods are demonstrated on real-world data where blind application of the models can lead to poor predictions and inference. Supplementary materials for this article are available online.

Experimental design and power calculation in omics circadian rhythmicity detection using the cosinor model.

Circadian clocks are 24-h endogenous oscillators in physiological and behavioral processes. Though recent transcriptomic studies have been successful in revealing the circadian rhythmicity in gene expression, the power calculation for omics circadian analysis have not been fully explored. In this paper, we develop a statistical method, namely CircaPower, to perform power calculation for circadian pattern detection. Our theoretical framework is determined by three key factors in circadian gene detection: sample size, intrinsic effect size and sampling design. Via simulations, we systematically investigate the impact of these key factors on circadian power calculation. We not only demonstrate that CircaPower is fast and accurate, but also show its underlying cosinor model is robust against variety of violations of model assumptions. In real applications, we demonstrate the performance of CircaPower using mouse pan-tissue data and human post-mortem brain data, and illustrate how to perform circadian power calculation using mouse skeleton muscle RNA-Seq pilot as case study. Our method CircaPower has been implemented in an R package, which is made publicly available on GitHub ( https://github.com/circaPower/circaPower).

Two-stage path analysis with definition variables: An alternative framework to account for measurement error.

When estimating path coefficients among psychological constructs measured with error, structural equation modeling (SEM), which simultaneously estimates the measurement and structural parameters, is generally regarded as the gold standard. In practice, however, researchers usually first compute composite scores or factor scores, and use those as observed variables in a path analysis, for purposes of simplifying the model or avoiding model convergence issues. Whereas recent approaches, such as reliability adjustment methods and factor score regression, has been proposed to mitigate the bias induced by ignoring measurement error in composite/factor scores with continuous indicators, those approaches are not yet applicable to models with categorical indicators. In this article, we introduce the two-stage path analysis (2S-PA) with definition variables as a general framework for path modeling to handle categorical indicators, in which estimation of factor scores and path coefficients are separated. It thus allows for different estimation methods in the measurement and the structural path models and easier diagnoses of violations of model assumptions. We conducted three simulation studies, ranging from latent regression to mediation analysis with categorical indicators, and showed that 2S-PA generally produced similar estimates to those using SEM in large samples, but gave better convergence rates, less standard error bias, and better control of Type I error rates in small samples. We illustrate 2S-PA using data from a national data set, and show how researchers can implement it in Mplus and OpenMx. Possible extensions and future directions of 2S-PA are discussed. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

The Causal Effects of Insomnia on Bipolar Disorder, Depression, and Schizophrenia: A Two-Sample Mendelian Randomization Study.

Psychiatric disorder, including bipolar disorder (BD), major depression (MDD), and schizophrenia (SCZ), affects millions of persons around the world. Understanding the disease causal mechanism underlying the three diseases and identifying the modifiable risk factors for them hold the key for the development of effective preventative and treatment strategies. We used a two-sample Mendelian randomization method to assess the causal effect of insomnia on the risk of BD, MDD, and SCZ in a European population. We collected one dataset of insomnia, three of BD, one of MDD, and three of SCZ and performed a meta-analysis for each trait, further verifying the analysis through extensive complementarity and sensitivity analysis. Among the three psychiatric disorders, we found that only insomnia is causally associated with MDD and that higher insomnia increases the risk of MDD. Specifically, the odds ratio of MDD increase of insomnia is estimated to be 1.408 [95% confidence interval (CI): 1.210–1.640, p = 1.03E-05] in the European population. The identified causal relationship between insomnia and MDD is robust with respect to the choice of statistical methods and is validated through extensive sensitivity analyses that guard against various model assumption violations. Our results provide new evidence to support the causal effect of insomnia on MDD and pave ways for reducing the psychiatric disorder burden.

Statistical Models for Estimating Linear Growth Velocity

Poor linear growth among infants is still a global public health issue. Linear growth velocity has been variously suggested as a more robust measure for growth over the classical measure of attained height for age. In this study, we systematically reviewed available literature for models used in estimating linear growth velocity. We searched Medline, Embase, Cochrane methodology register, Joanna Briggs Institute EBP, through the Ovid interface, and PubMed database to identify relevant articles that used statistical models to estimate linear growth velocity among infants. Longitudinal studies published in English were included. Two reviewers independently screened the titles and abstracts to identify potentially eligible studies. Any disagreements were discussed and resolved. Full-text articles were downloaded for all the studies that met the eligibility criteria. We synthesized literature using the preferred reporting items for systematic review and meta-analyses guidelines for the most used statistical methods for modelling infant growth trajectories. A total of 301 articles were retrieved from the initial search. Fifty-six full-text articles were assessed for eligibility and 16 of which were included in the final review with a total of 303,940 infants, median sample size of 732 (interquartile range: 241–1683). Polynomial function models were the most used growth model. Three (18.8%) of the articles modelled the linear growth. Two (12.5%) articles used mixed-effects models and another two (12.5%) used the Jenss-Bayley growth models to model linear growth. Other models included residual growth model, two-stage multilevel linear spline model, joint multilevel linear spline model, and generalized least squares with random effects. We have identified linear mixed-effects models, polynomial growth models, and the Jenss-Bayley model as the used models for characterizing linear growth among infants. Linear mixed-effects model is appealing for its robustness even under violation of largely robust even to quite severe violations of model assumptions.

An Attention-Based Diffusion Model for Psychometric Analyses

The emergence of computer-based assessments has made response times, in addition to response accuracies, available as a source of information about test takers’ latent abilities. The development of substantively meaningful accounts of the cognitive process underlying item responses is critical to establishing the validity of psychometric tests. However, existing substantive theories such as the diffusion model have been slow to gain traction due to their unwieldy functional form and regular violations of model assumptions in psychometric contexts. In the present work, we develop an attention-based diffusion model based on process assumptions that are appropriate for psychometric applications. This model is straightforward to analyse using Gibbs sampling and can be readily extended. We demonstrate our model’s good computational and statistical properties in a comparison with two well-established psychometric models.

A tutorial on Bayesian multi-model linear regression with BAS and JASP

Linear regression analyses commonly involve two consecutive stages of statistical inquiry. In the first stage, a single ‘best’ model is defined by a specific selection of relevant predictors; in the second stage, the regression coefficients of the winning model are used for prediction and for inference concerning the importance of the predictors. However, such second-stage inference ignores the model uncertainty from the first stage, resulting in overconfident parameter estimates that generalize poorly. These drawbacks can be overcome by model averaging, a technique that retains all models for inference, weighting each model’s contribution by its posterior probability. Although conceptually straightforward, model averaging is rarely used in applied research, possibly due to the lack of easily accessible software. To bridge the gap between theory and practice, we provide a tutorial on linear regression using Bayesian model averaging in JASP, based on the BAS package in R. Firstly, we provide theoretical background on linear regression, Bayesian inference, and Bayesian model averaging. Secondly, we demonstrate the method on an example data set from the World Happiness Report. Lastly, we discuss limitations of model averaging and directions for dealing with violations of model assumptions.

On classification with nonignorable missing data

We consider the problem of kernel classification with nonignorable missing data. Instead of imposing a fully parametric model for the selection probability, which can be quite sensitive to the violations of model assumptions, here we consider a semiparametric exponential tilting selection probability model in the spirit of Kim and Yu (2011). In addition to the existing parameter estimators, we also develop some new estimators of the unknown components of the model that are particularly suitable for classification problems. We also study various strong optimality properties of the proposed kernel-type classifiers.

A Bayesian State-Space Approach to Mapping Directional Brain Networks

The human brain is a directional network system of brain regions involving directional connectivity. Seizures are a directional network phenomenon as abnormal neuronal activities start from a seizure onset zone (SOZ) and propagate to otherwise healthy regions. To localize the SOZ of an epileptic patient, clinicians use intracranial electroencephalography (iEEG) to record the patient’s intracranial brain activity in many small regions. iEEG data are high-dimensional multivariate time series. We build a state-space multivariate autoregression (SSMAR) for iEEG data to model the underlying directional brain network. To produce scientifically interpretable network results, we incorporate into the SSMAR the scientific knowledge that the underlying brain network tends to have a cluster structure. Specifically, we assign to the SSMAR parameters a stochastic-blockmodel-motivated prior, which reflects the cluster structure. We develop a Bayesian framework to estimate the SSMAR, infer directional connections, and identify clusters for the unobserved network edges. The new method is robust to violations of model assumptions and outperforms existing network methods. By applying the new method to an epileptic patient’s iEEG data, we reveal seizure initiation and propagation in the patient’s directional brain network and discover a unique directional connectivity property of the SOZ. Overall, the network results obtained in this study bring new insights into epileptic patients’ normal and abnormal epileptic brain mechanisms and have the potential to assist neurologists and clinicians in localizing the SOZ—a long-standing research focus in epilepsy diagnosis and treatment. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.

Comparison of censoring assumptions to reduce bias in tuberculosis treatment cohort analyses.

ObjectiveObservational tuberculosis cohort studies are often limited by a lack of long-term data characterizing survival beyond the initial treatment outcome. Though Cox proportional hazards models are often applied to these data, differential risk of long-term survival, dependent on the initial treatment outcome, can lead to violations of model assumptions. We evaluate the performance of two alternate censoring approaches on reducing bias in treatment effect estimates.DesignWe simulate a typical multidrug-resistant tuberculosis cohort study and use Cox proportional hazards models to assess the relationship of an aggressive treatment regimen with hazard of death. We compare three assumptions regarding censored observations to determine which produces least biased treatment effect estimates: conventional non-informative censoring, an extension of short-term survival informed by literature, and incorporation of predicted long-term vital status.ResultsThe treatment regimen’s protective effect on death is consistently underestimated by the conventional censoring method, up to 7.6%. Models using the two alternative censoring techniques produce treatment effect estimates consistently stronger and less biased than the conventional method, underestimating the treatment effect by less than 2.4% across all scenarios.ConclusionsWhen model assumptions are violated, alternative censoring techniques can more accurately estimate associations between treatment and long-term survival. In multidrug-resistant tuberculosis cohort analyses, this bias reduction may yield more accurate and, larger effect estimates. This bias reduction can be achieved through use of standard statistical procedures with a simple re-coding of the censoring indicator.