HIERFSTAT, a package for R to compute and test hierarchical F-statistics

The package HIERFSTAT for the statistical software R, created by the R Development Core Team, allows the estimate of hierarchical F-statistics from a hierarchy with any numbers of levels. In addition, it allows testing the statistical significance of population differentiation for these different levels, using a generalized likelihood-ratio test. The package HIERFSTAT is available at http://www.unil.ch/popgen/softwares/hierfstat.htm.

Active-control trials with binary data: a comparison of methods for testing superiority or non-inferiority using the odds ratio

This paper considers methods for testing for superiority or non-inferiority in active-control trials with binary data, when the relative treatment effect is expressed as an odds ratio. Three asymptotic tests for the log-odds ratio based on the unconditional binary likelihood are presented, namely the likelihood ratio, Wald and score tests. All three tests can be implemented straightforwardly in standard statistical software packages, as can the corresponding confidence intervals. Simulations indicate that the three alternatives are similar in terms of the Type I error, with values close to the nominal level. However, when the non-inferiority margin becomes large, the score test slightly exceeds the nominal level. In general, the highest power is obtained from the score test, although all three tests are similar and the observed differences in power are not of practical importance. Copyright (C) 2007 John Wiley & Sons, Ltd.

Improved likelihood inference for the shape parameter in Weibull regression

We obtain adjustments to the profile likelihood function in Weibull regression models with and without censoring. Specifically, we consider two different modified profile likelihoods: (i) the one proposed by Cox and Reid [Cox, D.R. and Reid, N., 1987, Parameter orthogonality and approximate conditional inference. Journal of the Royal Statistical Society B, 49, 1-39.], and (ii) an approximation to the one proposed by Barndorff-Nielsen [Barndorff-Nielsen, O.E., 1983, On a formula for the distribution of the maximum likelihood estimator. Biometrika, 70, 343-365.], the approximation having been obtained using the results by Fraser and Reid [Fraser, D.A.S. and Reid, N., 1995, Ancillaries and third-order significance. Utilitas Mathematica, 47, 33-53.] and by Fraser et al. [Fraser, D.A.S., Reid, N. and Wu, J., 1999, A simple formula for tail probabilities for frequentist and Bayesian inference. Biometrika, 86, 655-661.]. We focus on point estimation and likelihood ratio tests on the shape parameter in the class of Weibull regression models. We derive some distributional properties of the different maximum likelihood estimators and likelihood ratio tests. The numerical evidence presented in the paper favors the approximation to Barndorff-Nielsen`s adjustment.

Improved likelihood inference in beta regression

We consider the issue of performing accurate small-sample likelihood-based inference in beta regression models, which are useful for modelling continuous proportions that are affected by independent variables. We derive small-sample adjustments to the likelihood ratio statistic in this class of models. The adjusted statistics can be easily implemented from standard statistical software. We present Monte Carlo simulations showing that inference based on the adjusted statistics we propose is much more reliable than that based on the usual likelihood ratio statistic. A real data example is presented.

Improved likelihood inference in Birnbaum-Saunders regressions

The Birnbaum-Saunders regression model is commonly used in reliability studies. We address the issue of performing inference in this class of models when the number of observations is small. Our simulation results suggest that the likelihood ratio test tends to be liberal when the sample size is small. We obtain a correction factor which reduces the size distortion of the test. Also, we consider a parametric bootstrap scheme to obtain improved critical values and improved p-values for the likelihood ratio test. The numerical results show that the modified tests are more reliable in finite samples than the usual likelihood ratio test. We also present an empirical application. (C) 2009 Elsevier B.V. All rights reserved.

INFERENCE FOR EIGENVALUES AND EIGENVECTORS OF GAUSSIAN SYMMETRIC MATRICES

This article presents maximum likelihood estimators (MLEs) and log-likelihood ratio (LLR) tests for the eigenvalues and eigenvectors of Gaussian random symmetric matrices of arbitrary dimension, where the observations are independent repeated samples from one or two populations. These inference problems are relevant in the analysis of diffusion tensor imaging data and polarized cosmic background radiation data, where the observations are, respectively, 3 x 3 and 2 x 2 symmetric positive definite matrices. The parameter sets involved in the inference problems for eigenvalues and eigenvectors are subsets of Euclidean space that are either affine subspaces, embedded submanifolds that are invariant under orthogonal transformations or polyhedral convex cones. We show that for a class of sets that includes the ones considered in this paper, the MLEs of the mean parameter do not depend on the covariance parameters if and only if the covariance structure is orthogonally invariant. Closed-form expressions for the MLEs and the associated LLRs are derived for this covariance structure.

A comparison of lipid variables as predictors of cardiovascular disease in the Asia Pacific Region

PURPOSE: Many guidelines advocate measurement of total or low density lipoprotein cholesterol (LDL), high density lipoprotein cholesterol (HDL), and triglycerides (TG) to determine treatment recommendations for preventing coronary heart disease (CHD) and cardiovascular disease (CVD). This analysis is a comparison of lipid variables as predictors of cardiovascular disease. METHODS: Hazard ratios for coronary and cardiovascular deaths by fourths of total cholesterol (TC), LDL, HDL, TG, non-HDL, TC/HDL, and TG/HDL values, and for a one standard deviation change in these variables, were derived in an individual participant data meta-analysis of 32 cohort studies conducted in the Asia-Pacific region. The predictive value of each lipid variable was assessed using the likelihood ratio statistic. RESULTS: Adjusting for confounders and regression dilution, each lipid variable had a positive (negative for HDL) log-linear association with fatal CHD and CVD. Individuals in the highest fourth of each lipid variable had approximately twice the risk of CHD compared with those with lowest levels. TG and HDL were each better predictors of CHD and CVD risk compared with TC alone, with test statistics similar to TC/HDL and TG/HDL ratios. Calculated LDL was a relatively poor predictor. CONCLUSIONS: While LDL reduction remains the main target of intervention for lipid-lowering, these data support the potential use of TG or lipid ratios for CHD risk prediction. (c) 2005 Elsevier Inc. All rights reserved.

Thoracic aorta calcification detected by electron beam tomography predicts all-cause mortality

Background: The presence of coronary artery calcium (CAC) is an independent marker of increased risk of cardiovascular disease (CVD) events and mortality. However, the predictive value of thoracic aorta calcification (TAC), which can be additionally identified without further scanning during assessment of CAC, is unknown. Methods: We followed a cohort of 8401 asymptomatic individuals (mean age: 53 +/- 10 years, 69% men) undergoing cardiac risk factor evaluation and TAC and CAC testing with electron beam computed tomography. Multivariable Cox proportional hazards models were developed to predict all-cause mortality based on the presence of TAC. Results: During a median follow-up period of 5 years, 124 (1.5%) deaths were observed. Overall survival was 96.9% and 98.9% for those with and without detectable TAC, respectively (p < 0.0001). Compared to those with no TAC, the hazard ratio for mortality in the presence of TAC was 3.25 (95% CI: 2.28-4.65, p < 0.0001) in unadjusted analysis. After adjusting for age, gender, hypertension, dyslipidemia, diabetes mellitus, smoking and family history of premature coronary artery disease, and presence of CAC the relationship remained robust (HR 1.61, 95% CI: 1.10-2.27, p = 0.015). Likelihood ratio chi(2) statistics demonstrated that the addition of TAC contributed significantly in predicting mortality to traditional risk factors alone (chi(2) = 13.62, p = 0.002) as well as risk factors + CAC (chi(2) = 5.84, p = 0.02) models. Conclusion: In conclusion, the presence of TAC was associated with all-cause mortality in our study; this relationship was independent of conventional CVD risk factors as well as the presence of CAC. (C) 2009 Elsevier Ireland Ltd. All rights reserved.

Relation of Aortic Valve Calcium Detected by Cardiac Computed Tomography to All-Cause Mortality

Aortic valve calcium (AVC) can be quantified on the same computed tomographic scan as coronary artery calcium (CAC). Although CAC is an established predictor of cardiovascular events, limited evidence is available for an independent predictive value for AVC. We studied a cohort of 8,401 asymptomatic subjects (mean age 53 10 years, 69% men), who were free of known coronary heart disease and were undergoing electron beam computed tomography for assessment of subclinical atherosclerosis. The patients were followed for a median of 5 years (range 1 to 7) for the occurrence of mortality from any cause. Multivariate Cox regression models were developed to predict all-cause mortality according to the presence of AVC. A total of 517 patients (6%) had AVC on electron beam computed tomography. During follow-up, 124 patients died (1.5%), for an overall survival rate of 96.1% and 98.7% for those with and without AVC, respectively (hazard ratio 3.39, 95% confidence interval 2.09 to 5.49). After adjustment for age, gender, hypertension, dyslipidemia, diabetes mellitus, smoking, and a family history of premature coronary heart disease, AVC remained a significant predictor of mortality (hazard ratio 1.82, 95% confidence interval 1.11 to 2.98). Likelihood ratio chi-square statistics demonstrated that the addition of AVC contributed significantly to the prediction of mortality in a model adjusted for traditional risk factors (chi-square = 5.03, p = 0.03) as well as traditional risk factors plus the presence of CAC (chi-square = 3.58, p = 0.05). In conclusion, AVC was associated with increased all-cause mortality, independent of the traditional risk factors and the presence of CAC. (C) 2010 Published by Elsevier Inc. (Am J Cardiol 2010;106:1787-1791)

Modelling the distribution of stamp paper thickness via finite normal mixtures: The 1872 Hidalgo stamp issue of Mexico revisited

Izenman and Sommer (1988) used a non-parametric Kernel density estimation technique to fit a seven-component model to the paper thickness of the 1872 Hidalgo stamp issue of Mexico. They observed an apparent conflict when fitting a normal mixture model with three components with unequal variances. This conflict is examined further by investigating the most appropriate number of components when fitting a normal mixture of components with equal variances.

A mixture model-based approach to the clustering of microarray expression data

Motivation: This paper introduces the software EMMIX-GENE that has been developed for the specific purpose of a model-based approach to the clustering of microarray expression data, in particular, of tissue samples on a very large number of genes. The latter is a nonstandard problem in parametric cluster analysis because the dimension of the feature space (the number of genes) is typically much greater than the number of tissues. A feasible approach is provided by first selecting a subset of the genes relevant for the clustering of the tissue samples by fitting mixtures of t distributions to rank the genes in order of increasing size of the likelihood ratio statistic for the test of one versus two components in the mixture model. The imposition of a threshold on the likelihood ratio statistic used in conjunction with a threshold on the size of a cluster allows the selection of a relevant set of genes. However, even this reduced set of genes will usually be too large for a normal mixture model to be fitted directly to the tissues, and so the use of mixtures of factor analyzers is exploited to reduce effectively the dimension of the feature space of genes. Results: The usefulness of the EMMIX-GENE approach for the clustering of tissue samples is demonstrated on two well-known data sets on colon and leukaemia tissues. For both data sets, relevant subsets of the genes are able to be selected that reveal interesting clusterings of the tissues that are either consistent with the external classification of the tissues or with background and biological knowledge of these sets.

A note on quasi-stationary distributions of birth-death processes and the SIS logistic epidemic

For Markov processes on the positive integers with the origin as an absorbing state, Ferrari, Kesten, Martinez and Picco studied the existence of quasi-stationary and limiting conditional distributions by characterizing quasi-stationary distributions as fixed points of a transformation Phi on the space of probability distributions on {1, 2,.. }. In the case of a birth-death process, the components of Phi(nu) can be written down explicitly for any given distribution nu. Using this explicit representation, we will show that Phi preserves likelihood ratio ordering between distributions. A conjecture of Kryscio and Lefevre concerning the quasi-stationary distribution of the SIS logistic epidemic follows as a corollary.

New perspectives in the use of ink evidence in forensic science: Part I. Development of a quality assurance process for forensic ink analysis by HPTLC

The level of information provided by ink evidence to the criminal and civil justice system is limited. The limitations arise from the weakness of the interpretative framework currently used, as proposed in the ASTM 1422-05 and 1789-04 on ink analysis. It is proposed to use the likelihood ratio from the Bayes theorem to interpret ink evidence. Unfortunately, when considering the analytical practices, as defined in the ASTM standards on ink analysis, it appears that current ink analytical practices do not allow for the level of reproducibility and accuracy required by a probabilistic framework. Such framework relies on the evaluation of the statistics of the ink characteristics using an ink reference database and the objective measurement of similarities between ink samples. A complete research programme was designed to (a) develop a standard methodology for analysing ink samples in a more reproducible way, (b) comparing automatically and objectively ink samples and (c) evaluate the proposed methodology in a forensic context. This report focuses on the first of the three stages. A calibration process, based on a standard dye ladder, is proposed to improve the reproducibility of ink analysis by HPTLC, when these inks are analysed at different times and/or by different examiners. The impact of this process on the variability between the repetitive analyses of ink samples in various conditions is studied. The results show significant improvements in the reproducibility of ink analysis compared to traditional calibration methods.

Hardy-Weinberg Equilibrium and the Ternary Plot

The Hardy-Weinberg law, formulated about 100 years ago, states that under certainassumptions, the three genotypes AA, AB and BB at a bi-allelic locus are expected to occur inthe proportions p2, 2pq, and q2 respectively, where p is the allele frequency of A, and q = 1-p.There are many statistical tests being used to check whether empirical marker data obeys theHardy-Weinberg principle. Among these are the classical xi-square test (with or withoutcontinuity correction), the likelihood ratio test, Fisher's Exact test, and exact tests in combinationwith Monte Carlo and Markov Chain algorithms. Tests for Hardy-Weinberg equilibrium (HWE)are numerical in nature, requiring the computation of a test statistic and a p-value.There is however, ample space for the use of graphics in HWE tests, in particular for the ternaryplot. Nowadays, many genetical studies are using genetical markers known as SingleNucleotide Polymorphisms (SNPs). SNP data comes in the form of counts, but from the countsone typically computes genotype frequencies and allele frequencies. These frequencies satisfythe unit-sum constraint, and their analysis therefore falls within the realm of compositional dataanalysis (Aitchison, 1986). SNPs are usually bi-allelic, which implies that the genotypefrequencies can be adequately represented in a ternary plot. Compositions that are in exactHWE describe a parabola in the ternary plot. Compositions for which HWE cannot be rejected ina statistical test are typically “close" to the parabola, whereas compositions that differsignificantly from HWE are “far". By rewriting the statistics used to test for HWE in terms ofheterozygote frequencies, acceptance regions for HWE can be obtained that can be depicted inthe ternary plot. This way, compositions can be tested for HWE purely on the basis of theirposition in the ternary plot (Graffelman & Morales, 2008). This leads to nice graphicalrepresentations where large numbers of SNPs can be tested for HWE in a single graph. Severalexamples of graphical tests for HWE (implemented in R software), will be shown, using SNPdata from different human populations

Simulation-Based Finite and Large Sample Tests in Multivariate Regressions.

In the context of multivariate linear regression (MLR) models, it is well known that commonly employed asymptotic test criteria are seriously biased towards overrejection. In this paper, we propose a general method for constructing exact tests of possibly nonlinear hypotheses on the coefficients of MLR systems. For the case of uniform linear hypotheses, we present exact distributional invariance results concerning several standard test criteria. These include Wilks' likelihood ratio (LR) criterion as well as trace and maximum root criteria. The normality assumption is not necessary for most of the results to hold. Implications for inference are two-fold. First, invariance to nuisance parameters entails that the technique of Monte Carlo tests can be applied on all these statistics to obtain exact tests of uniform linear hypotheses. Second, the invariance property of the latter statistic is exploited to derive general nuisance-parameter-free bounds on the distribution of the LR statistic for arbitrary hypotheses. Even though it may be difficult to compute these bounds analytically, they can easily be simulated, hence yielding exact bounds Monte Carlo tests. Illustrative simulation experiments show that the bounds are sufficiently tight to provide conclusive results with a high probability. Our findings illustrate the value of the bounds as a tool to be used in conjunction with more traditional simulation-based test methods (e.g., the parametric bootstrap) which may be applied when the bounds are not conclusive.