Comparative efficacy of three Bayesian variable selection methods in the context of weight loss in obese women.

The use of high-dimensional data has expanded in many fields, including in clinical research, thus making variable selection methods increasingly important compared to traditional statistical approaches. The work aims to compare the performance of three supervised Bayesian variable selection methods to detect the most important predictors among a high-dimensional set of variables and to provide useful and practical guidelines of their use. We assessed the variable selection ability of: (1) Bayesian Kernel Machine Regression (BKMR), (2) Bayesian Semiparametric Regression (BSR), and (3) Bayesian Least Absolute Shrinkage and Selection Operator (BLASSO) regression on simulated data of different dimensions and under three scenarios with disparate predictor-response relationships and correlations among predictors. This is the first study describing when one model should be preferred over the others and when methods achieve comparable results. BKMR outperformed all other models with small synthetic datasets. BSR was strongly dependent on the choice of its own intrinsic parameter, but its performance was comparable to BKMR with large datasets. BLASSO should be preferred only when it is reasonable to hypothesise the absence of synergies between predictors and the presence of monotonous predictor-outcome relationships. Finally, we applied the models to a real case study and assessed the relationships among anthropometric, biochemical, metabolic, cardiovascular, and inflammatory variables with weight loss in 755 hospitalised obese women from the Follow Up OBese patients at AUXOlogico institute (FUOBAUXO) cohort.

Beyond p < .05: a critical review of new Bayesian proposals for assessing the p-value.

This paper reviews recent contributions from a Bayesian-oriented perspective, after the ASA statement on p-values (2016). We classify proposals that (i) supplement the p-value; (ii) modify the p-value itself. In the first group, we review the Bayes factor, the False Positive risk, the rejection odds and the analysis of credibility from both Matthews' and Held's point of view. We also put forth and discuss a new index of credibility, about which we conduct a delimited simulation study. In the second group, we discuss Gannon's modification of the p-value based on the Bayes factor and the second-generation p-value. The theory is illustrated with two case studies on pharmacotherapy in infectious diseases. Contemporary authors still refer to the p-value as a statistical indicator but have abandoned the perspective of evaluating p-values with fixed thresholds. Statistical societies worldwide should target new strategies to disseminate the debate on p-values in all applied fields of knowledge, as well as they may promote the use of different statistical procedures to supplement p-values.

Evaluation of Inter-Observer Reliability of Animal Welfare Indicators: Which Is the Best Index to Use?

This study focuses on the problem of assessing inter-observer reliability (IOR) in the case of dichotomous categorical animal-based welfare indicators and the presence of two observers. Based on observations obtained from Animal Welfare Indicators (AWIN) project surveys conducted on nine dairy goat farms, and using udder asymmetry as an indicator, we compared the performance of the most popular agreement indexes available in the literature: Scott's π, Cohen's k, kPABAK, Holsti's H, Krippendorff's α, Hubert's Γ, Janson and Vegelius' J, Bangdiwala's B, Andrés and Marzo's ∆, and Gwet's γ(AC1). Confidence intervals were calculated using closed formulas of variance estimates for π, k, kPABAK, H, α, Γ, J, ∆, and γ(AC1), while the bootstrap and exact bootstrap methods were used for all the indexes. All the indexes and closed formulas of variance estimates were calculated using Microsoft Excel. The bootstrap method was performed with R software, while the exact bootstrap method was performed with SAS software. k, π, and α exhibited a paradoxical behavior, showing unacceptably low values even in the presence of very high concordance rates. B and γ(AC1) showed values very close to the concordance rate, independently of its value. Both bootstrap and exact bootstrap methods turned out to be simpler compared to the implementation of closed variance formulas and provided effective confidence intervals for all the considered indexes. The best approach for measuring IOR in these cases is the use of B or γ(AC1), with bootstrap or exact bootstrap methods for confidence interval calculation.

Best uses of p-values and complementary measures in medical research: Recent developments in the frequentist and Bayesian frameworks.

The p-value is a classical proposal of statistical inference, dating back to the seminal contributions by Fisher, Neyman and E. Pearson. However, p-values have been frequently misunderstood and misused in practice, and medical research is not an exception. In recent years, in several statistical and applied journals, a debate erupted about the need of clear guidelines in reporting p-values, which culminated with the publication of the ASA statement in 2016. In this paper, we assess strengths and limitations of p-values and we assert that in applied research the p-value should be supplemented by other measures, such as the Bayes factor, the Bayes false discovery rate and the local Bayes false discovery rate. We also review a recent proposal by Bayarri et al. from a Bayesian perspective that has the advantage of introducing an indicator, the rejection odds, which keeps into account both pre- and post-experimental information, and could also have a straightforward frequentist interpretation. We conduct a delimited numerical study that investigates on the relation of the Bayes factor with its maximum, and of the local Bayes false discovery rate with its minimum under different distributional assumptions and parameter choices. We illustrate the concepts expressed in theory with an example in clinical oncology, namely a randomized trial on the effectiveness of a new chemotherapy for patients with AIDS and Kaposi's sarcoma.

Brain networks construction using Bayes FDR and average power function.

Brain functional connectivity is a widely investigated topic in neuroscience. In recent years, the study of brain connectivity has been largely aided by graph theory. The link between time series recorded at multiple locations in the brain and the construction of a graph is usually an adjacency matrix. The latter converts a measure of the connectivity between two time series, typically a correlation coefficient, into a binary choice on whether the two brain locations are functionally connected or not. As a result, the choice of a threshold τ over the correlation coefficient is key. In the present work, we propose a multiple testing approach to the choice of τ that uses the Bayes false discovery rate and a new estimator of the statistical power called average power function to balance the two types of statistical error. We show that the proposed average power function estimator behaves well both in case of independence and weak dependence of the tests and it is reliable under several simulated dependence conditions. Moreover, we propose a robust method for the choice of τ using the 5% and 95% percentiles of the average power function and False Discovery Rate bootstrap distributions, respectively, to improve stability. We applied our approach to functional magnetic resonance imaging and high density electroencephalogram data.

Big data, observational research and P-value: a recipe for false-positive findings? A study of simulated and real prospective cohorts.

BACKGROUND: An increasing number of observational studies combine large sample sizes with low participation rates, which could lead to standard inference failing to control the false-discovery rate. We investigated if the 'empirical calibration of P-value' method (EPCV), reliant on negative controls, can preserve type I error in the context of survival analysis. METHODS: We used simulated cohort studies with 50% participation rate and two different selection bias mechanisms, and a real-life application on predictors of cancer mortality using data from four population-based cohorts in Northern Italy (n = 6976 men and women aged 25-74 years at baseline and 17 years of median follow-up). RESULTS: Type I error for the standard Cox model was above the 5% nominal level in 15 out of 16 simulated settings; for n = 10 000, the chances of a null association with hazard ratio = 1.05 having a P-value < 0.05 were 42.5%. Conversely, EPCV with 10 negative controls preserved the 5% nominal level in all the simulation settings, reducing bias in the point estimate by 80-90% when its main assumption was verified. In the real case, 15 out of 21 (71%) blood markers with no association with cancer mortality according to literature had a P-value < 0.05 in age- and gender-adjusted Cox models. After calibration, only 1 (4.8%) remained statistically significant. CONCLUSIONS: In the analyses of large observational studies prone to selection bias, the use of empirical distribution to calibrate P-values can substantially reduce the number of trivial results needing further screening for relevance and external validity.

Fuzzy Analysis of Students' Ratings.

BACKGROUND: Intuitionistic fuzzy sets (IFS) represent a methodology for quantifying latent variables in questionnaire analysis through membership and non-membership functions, which are linked by an uncertainty function. OBJECTIVES: We aim to apply an IFS approach to the problem of students' satisfaction of university teaching. Such framework can take into account a source of uncertainty related to items and another related to subjects. RESULTS: A new technique for IFS analysis is set forth and generalized to a multivariate scenario. Potential advantages of the IFS perspective with respect to other nonfuzzy approaches are provided. APPLICATION: We apply this method to a national program of university courses evaluation and we focus, in particular, on the outcomes of two Masters in Statistics.

Assessing the inter-rater agreement for ordinal data through weighted indexes.

Assessing the inter-rater agreement between observers, in the case of ordinal variables, is an important issue in both the statistical theory and biomedical applications. Typically, this problem has been dealt with the use of Cohen's weighted kappa, which is a modification of the original kappa statistic, proposed for nominal variables in the case of two observers. Fleiss (1971) put forth a generalization of kappa in the case of multiple observers, but both Cohen's and Fleiss' kappa could have a paradoxical behavior, which may lead to a difficult interpretation of their magnitude. In this paper, a modification of Fleiss' kappa, not affected by paradoxes, is proposed, and subsequently generalized to the case of ordinal variables. Monte Carlo simulations are used both to testing statistical hypotheses and to calculating percentile and bootstrap-t confidence intervals based on this statistic. The normal asymptotic distribution of the proposed statistic is demonstrated. Our results are applied to the classical Holmquist et al.'s (1967) dataset on the classification, by multiple observers, of carcinoma in situ of the uterine cervix. Finally, we generalize the use of s* to a bivariate case.

The hooded crow (Corvus cornix) as an environmental bioindicator species of heavy metal contamination.

This study aims to examine the possible presence of lead and cadmium in the liver and kidneys of hooded crows (Corvus cornix). Liver and kidneys of hooded crow carcasses were collected in Province of Cuneo (Piedmont, Italy) in order to detect lead and cadmium content. Significant differences were found in lead and cadmium levels between areas of intensive cultivation versus areas where meadows are prevalent. Moreover, age greatly influenced the burden of heavy metals, while sex did not seem to affect the level of contamination. The source of contamination may be phosphate fertilizers used for intensive cultivation in the study area.

pFDR and pFNR estimation for brain networks construction.

Recent developments in the study of brain functional connectivity are widely based on graph theory. In the current analysis of brain networks, there is no unique way to derive the adjacency matrix, which is a useful representation for a graph. Its entries, containing information about the existence of links, are identified by thresholding the correlation between the time series that characterized the dynamic behavior of the nodes. In this work, we put forward a strategy to choose a suitable threshold on the correlation matrix considering the problem of multiple comparisons in order to control the error rates. In this context we propose to control the positive false discovery rate (pFDR) and a similar measure involving false negatives, called the positive false nondiscovery rate (pFNR). In particular, we provide point and interval estimators for pFNR and a method for balancing the two types of error, demonstrating it by using functional magnetic resonance imaging data and Monte Carlo simulations.

Brain network connectivity assessed using graph theory in frontotemporal dementia.

OBJECTIVE: To investigate whether brain functional network connectivity is disrupted in patients with the behavioral variant of frontotemporal dementia (bvFTD). METHODS: Graph theoretical analysis was applied to resting state functional MRI data from 18 patients with probable bvFTD and 50 healthy individuals. Functional connectivity between 90 cortical and subcortical brain regions was estimated using bivariate correlation analysis and thresholded to construct a set of undirected graphs. Correlations between network properties and cognitive variables were tested. RESULTS: Global topologic organization of the functional brain network in bvFTD was significantly disrupted as indicated by reduced mean network degree, clustering coefficient, and global efficiency and increased characteristic path length and assortativity relative to normal subjects. Compared to controls, bvFTD data showed retention of major "hub" regions in the medial parietal, temporal, and occipital lobes, but cortical hubs were not noted in the frontal lobes. Medial and dorsal frontal regions, left caudate nucleus, left insular cortices, and some regions of the temporal, parietal, and occipital lobes showed decreased nodal centrality. BvFTD patients showed the greatest decrease in inter-regional connectivity between the frontal and occipital regions, and the insular cortices and occipital, temporal, subcortical, and frontal regions. In bvFTD, altered global network properties correlated with executive dysfunction. CONCLUSIONS: Global and local functional networks are altered in bvFTD, suggesting a loss of efficiency in information exchange between both distant and close brain areas. Altered brain regions are located in structures that are closely associated with neuropathologic changes in bvFTD. Aberrant topology of the functional brain networks in bvFTD appears to underlie cognitive deficits in these patients.

Flexible meta-regression functions for modeling aggregate dose-response data, with an application to alcohol and mortality.

In this paper, the authors describe fractional polynomials and cubic splines with which to represent smooth dose-response relations in summarizing meta-analytical aggregate data. Use of these two curve-fitting families can help prevent the problems arising from inappropriate linearity assumptions. These methods are illustrated in the problem of estimating the shape of the dose-response curve between alcohol consumption and all-cause mortality risk. The authors considered aggregate data from 29 cohort studies investigating this issue (1966-2000). J-shaped curves with a nadir at approximately 5-7 g/day of alcohol consumption and a last protective dose of 47-60 g/day were consistently obtained from fractional polynomials and cubic splines. The authors conclude that both of the curve-fitting families are useful tools with which to explore dose-response epidemiologic questions by means of meta-analytical approaches, especially when important nonlinearity is anticipated.