Statistical evidence in contingency tables analysis

The likelihood ratio is used for measuring the strength of statistical evidence. The probability of observing strong misleading evidence along with that of observing weak evidence evaluate the performance of this measure. When the corresponding likelihood function is expressed in terms of a parametric statistical model that fails, the likelihood ratio retains its evidential value if the likelihood function is robust [Royall, R., Tsou, T.S., 2003. Interpreting statistical evidence by using imperfect models: robust adjusted likelihood functions. J. Roy. Statist. Soc. Ser. B 65, 391–404]. In this paper, we extend the theory of Royall and Tsou [2003. Interpreting statistical evidence by using imperfect models: robust adjusted likelihood functions. J. Roy. Statist. Soc., Ser. B 65, 391–404] to the case when the assumed working model is a characteristic model for two-way contingency tables (the model of independence, association and correlation models). We observe that association and correlation models are not equivalent in terms of statistical evidence. The association models are bounded by the maximum of the bump function while the correlation models are not.     

Mixture Models in View of Evidential Analysis

In some practical inferential situations, it is needed to mix some finite sort of distributions to fit an adequate model for multi-modal observations. In this article, using evidential analysis, we determine the sample size for supporting hypotheses about the mixture proportion and homogeneity. An Expectation-Maximization algorithm is used to evaluate the probability of strong misleading evidence based on modified likelihood ratio as a measure of support.     

How Often Likelihood Ratios are Misleading in Sequential Trials

How often would investigators be misled if they took advantage of the likelihood principle and used likelihood ratios—which need not be adjusted for multiple looks at the data—to frequently examine accumulating data? The answer, perhaps surprisingly, is not often. As expected, the probability of observing misleading evidence does increase with each additional examination. However, the amount by which this probability increases converges to zero as the sample size grows. As a result, the probability of observing misleading evidence remains bounded—and therefore controllable—even with an infinite number of looks at the data. Here we use boundary crossing results to detail how often misleading likelihood ratios arise in sequential designs. We find that the probability of observing a misleading likelihood ratio is often much less than its universal bound. Additionally, we find that in the presence of fixed-dimensional nuisance parameters, profile likelihoods are to be preferred over estimated likelihoods which result from replacing the nuisance parameters by their global maximum likelihood estimates.     

Interpreting statistical evidence by using imperfect models: robust adjusted likelihood functions

Summary. The strength of statistical evidence is measured by the likelihood ratio. Two key performance properties of this measure are the probability of observing strong misleading evidence and the probability of observing weak evidence. For the likelihood function associated with a parametric statistical model, these probabilities have a simple large sample structure when the model is correct. Here we examine how that structure changes when the model fails. This leads to criteria for determining whether a given likelihood function is robust (continuing to perform satisfactorily when the model fails), and to a simple technique for adjusting both likelihoods and profile likelihoods to make them robust. We prove that the expected information in the robust adjusted likelihood cannot exceed the expected information in the likelihood function from a true model. We note that the robust adjusted likelihood is asymptotically fully efficient when the working model is correct, and we show that in some important examples this efficiency is retained even when the working model fails. In such cases the Bayes posterior probability distribution based on the adjusted likelihood is robust, remaining correct asymptotically even when the model for the observable random variable does not include the true distribution. Finally we note a link to standard frequentist methodology—in large samples the adjusted likelihood functions provide robust likelihood-based confidence intervals.     

Axiomatic Development of Profile Likelihoods as the Strength of Evidence for Composite Hypotheses

One important type of question in statistical inference is how to interpret data as evidence. The law of likelihood provides a satisfactory answer in interpreting data as evidence for simple hypotheses, but remains silent for composite hypotheses. This article examines how the law of likelihood can be extended to composite hypotheses within the scope of the likelihood principle. From a system of axioms, we conclude that the strength of evidence for the composite hypotheses should be represented by an interval between lower and upper profiles likelihoods. This article is intended to reveal the connection between profile likelihoods and the law of likelihood under the likelihood principle rather than argue in favor of the use of profile likelihoods in addressing general questions of statistical inference. The interpretation of the result is also discussed.     

Model fusion and multiple testing in the likelihood paradigm: shrinkage and evidence supporting a point null hypothesis

ABSTRACT

According to the general law of likelihood, the strength of statistical evidence for a hypothesis as opposed to its alternative is the ratio of their likelihoods, each maximized over the parameter of interest. Consider the problem of assessing the weight of evidence for each of several hypotheses. Under a realistic model with a free parameter for each alternative hypothesis, this leads to weighing evidence without any shrinkage toward a presumption of the truth of each null hypothesis. That lack of shrinkage can lead to many false positives in settings with large numbers of hypotheses. A related problem is that point hypotheses cannot have more support than their alternatives. Both problems may be solved by fusing the realistic model with a model of a more restricted parameter space for use with the general law of likelihood. Applying the proposed framework of model fusion to data sets from genomics and education yields intuitively reasonable weights of evidence.     

Comparison of record data and random observations based on statistical evidence

“Science looks to Statistics for an objective measure of the strength of evidence in a given body of observations. The Law of Likelihood explains that the strength of statistical evidence for one hypothesis over another is measured by their likelihood ratio” (Blume, 2002). In this paper, we compare probabilities of weak and strong misleading evidence based on record data with those based on the same number of iid observations from the original distribution. We shall also use a criterion defined as a combination of probabilities of weak and strong misleading evidence to do the above comparison. We also give numerical results of a simulated comparison.     

Statistical Evidence in Experiments and in Record Values

According to the law of likelihood, statistical evidence for one (simple) hypothesis against another is measured by their likelihood ratio. When the experimenter can choose between two or more experiments (of approximately the same cost) to obtain data, he would want to know which experiment provides (on average) stronger true evidence for one hypothesis against another. In this article, after defining a pre-experimental criterion for the potential strength of evidence provided by an experiment, based on entropy distance, we compare the potential statistical evidence in lower record values with that in the same number of iid observations from the same parent distribution. We also establish a relation between Fisher information and Kullback–Leibler distance.     

Likelihood ratio tests for fuzzy hypotheses testing

In hypotheses testing, such as other statistical problems, we may confront imprecise concepts. One case is a situation in which hypotheses are imprecise. In this paper, we recall and redefine some concepts about fuzzy hypotheses testing, and then we introduce the likelihood ratio test for fuzzy hypotheses testing. Finally, we give some applied examples.     

Two-stage informative cluster sampling—estimation and prediction with applications for small-area models

This paper considers the effects of informative two-stage cluster sampling on estimation and prediction. The aims of this article are twofold: first to estimate the parameters of the superpopulation model for two-stage cluster sampling from a finite population, when the sampling design for both stages is informative, using maximum likelihood estimation methods based on the sample-likelihood function; secondly to predict the finite population total and to predict the cluster-specific effects and the cluster totals for clusters in the sample and for clusters not in the sample. To achieve this we derive the sample and sample-complement distributions and the moments of the first and second stage measurements. Also we derive the conditional sample and conditional sample-complement distributions and the moments of the cluster-specific effects given the cluster measurements. It should be noted that classical design-based inference that consists of weighting the sample observations by the inverse of sample selection probabilities cannot be applied for the prediction of the cluster-specific effects for clusters not in the sample. Also we give an alternative justification of the Royall [1976. The linear least squares prediction approach to two-stage sampling. Journal of the American Statistical Association 71, 657–664] predictor of the finite population total under two-stage cluster population. Furthermore, small-area models are studied under informative sampling.     

Evidential inference based on record data and inter-record times

Science looks to statistics for an objective measure of the strength of evidence in a given body of observations. In this paper, we shall use a criterion defined as a combination of probabilities of weak and strong misleading evidence to do the comparison between only record values and the same number of record values and inter-record times. Also, a simulation is presented to illustrate the results.     

Estimates for mean and dispersion effects in unreplicated factorial designs

One important goal of experimentation in quality improvement is to minimize the variability of a product or process around a target mean value. Factors which affect variances as well as factors that affect the mean can be identified using the analysis of mean and dispersion. Box and Meyer (1986b) proposed a method of model identification and maximum likelihood estimation for mean and dispersion effects from unreplicated designs. In this article, we address two problems associated with MLE’s. First, asymptotic variance of MLE's for dispersion effects which can be used to judge the significance of factors can be misleading. A possible explanation is provided; simulation results also indicate that the asymptotic, variance underestimates.     

Applying skovgaard's modified directed likelihood statistic to mixed linear models

The introduction of software to calculate maximum likelihood estimates for mixed linear models has made likelihood estimation a practical alternative to methods based on sums of squares. Likelihood based tests and confidence intervals, however, may be misleading in problems with small sample sizes. This paper discusses an adjusted version of the directed log-likelihood statistic for mixed models that is highly accurate for testing one parameter hypotheses. Indroduced by Skovgaard (1996, Journal of the Bernoulli Society,2,145-165), we show in mixed models that the statistic has a simple conpact from that may be obtained from standard software. Simulation studies indicate that this statistic is more accurate than many of the specialized procedure that have been advocated.     

On limiting likelihood ratio processes of some change-point type statistical models

Different change-point type models encountered in parametric statistical inference give rise to different limiting likelihood ratio processes. In this paper we consider two such likelihood ratios. The first one is an exponential functional of a two-sided Poisson process driven by some parameter, while the second one is an exponential functional of a two-sided Brownian motion. We establish that for sufficiently small values of the parameter, the Poisson type likelihood ratio can be approximated by the Brownian type one. As a consequence, several statistically interesting quantities (such as limiting variances of different estimators) related to the first likelihood ratio can also be approximated by those related to the second one. Finally, we discuss the asymptotics for large values of the parameter and illustrate the results by numerical simulations.     

A self-adjusted weighted likelihood ratio test for global clustering of disease

Compared to tests for localized clusters, the tests for global clustering only collect evidence for clustering throughout the study region without evaluating the statistical significance of the individual clusters. The weighted likelihood ratio (WLR) test based on the weighted sum of likelihood ratios represents an important class of tests for global clustering. Song and Kulldorff (Likelihood based tests for spatial randomness. Stat Med. 2006;25(5):825–839) developed a wide variety of weight functions with the WLR test for global clustering. However, these weight functions are often defined based on the cell population size or the geographic information such as area size and distance between cells. They do not make use of the information from the observed count, although the likelihood ratio of a potential cluster depends on both the observed count and its population size. In this paper, we develop a self-adjusted weight function to directly allocate weights onto the likelihood ratios according to their values. The power of the test was evaluated and compared with existing methods based on a benchmark data set. The comparison results favour the suggested test especially under global chain clustering models.     

Likelihood Asymptotics

The paper gives an overview of modern likelihood asymptotics with emphasis on results and applicability. Only parametric inference in well-behaved models is considered and the theory discussed leads to highly accurate asymptotic tests for general smooth hypotheses. The tests are refinements of the usual asymptotic likelihood ratio tests, and for one-dimensional hypotheses the test statistic is known as r *, introduced by Barndorff-Nielsen. Examples illustrate the applicability and accuracy as well as the complexity of the required computations. Modern likelihood asymptotics has developed by merging two lines of research: asymptotic ancillarity is the basis of the statistical development, and saddlepoint approximations or Laplace-type approximations have simultaneously developed as the technical foundation. The main results and techniques of these two lines will be reviewed, and a generalization to multi-dimensional tests is developed. In the final part of the paper further problems and ideas are presented. Among these are linear models with non-normal error, non-parametric linear models obtained by estimation of the residual density in combination with the present results, and the generalization of the results to restricted maximum likelihood and similar structured models.     

Weighing the Evidence for Hypotheses with Small Samples of Right-censored Exponential Data

The p-value evidence for an alternative to a null hypothesis regarding the mean lifetime can be unreliable if based on asymptotic approximations when there is only a small sample of right-censored exponential data. However, a guarded weight of evidence for the alternative can always be obtained without approximation, no matter how small the sample, and has some other advantages over p-values. Weights of evidence are defined as estimators of 0 when the null hypothesis is true and 1 when the alternative is true, and they are judged on the basis of the ensuing risks, where risk is mean squared error of estimation. The evidence is guarded in that a preassigned bound is placed on the risk under the hypothesis. Practical suggestions are given for choosing the bound and for interpreting the magnitude of the weight of evidence. Acceptability profiles are obtained by inversion of a family of guarded weights of evidence for two-sided alternatives to point hypotheses, just as confidence intervals are obtained from tests; these profiles are arguably more informative than confidence intervals, and are easily determined for any level and any sample size, however small. They can help understand the effects of different amounts of censoring. They are found for several small size data sets, including a sample of size 12 for post-operative cancer patients. Both singly Type I and Type II censored examples are included. An examination of the risk functions of these guarded weights of evidence suggests that if the censoring time is of the same magnitude as the mean lifetime, or larger, then the risks in using a guarded weight of evidence based on a likelihood ratio are not much larger than they would be if the parameter were known.     

An Estimation Method of the Pair Potential Function for Gibbs Point Processes on Spheres

An estimation method for pairwise interaction potential of a stationary Gibbs point process is introduced by considering the case of observations located on a sphere. It is based both on Fourier decomposition of the potential and on minimum contrast estimation. It is defined when many independent realizations of the process are available. Consistency and asymptotic normality are proved for the resulting estimators. The method enables derivation of the choice of the potential function by embedded hypotheses testing. The method is applied to independent observations of root locations on internodes around stem of maize roots. The internodes are described as circles and we focus on the interaction function associated with the potential. Since a model with too many components seems to fail, we choose a sequential procedure based on embedded hypotheses testing to build a simpler model.     

Testing hypotheses in truncated samples by means of divergence statistics

In this paper we consider the problem of testing hypotheses in parametric models, when only the first r (of n) ordered observations are known.Using divergence measures, a procedure to test statistical hypotheses is proposed, Replacing the parameters by suitable estimators in the expresion of the divergence measure, the test statistics are obtained.Asymptotic distributions for these statistics are given in several cases when maximum likelihood estimators for truncated samples are considered.Applications of these results in testing statistical hypotheses, on the basis of truncated data, are presented.The small sample behavior of the proposed test statistics is analyzed in particular cases.A comparative study of power values is carried out by computer simulation.     

Estimating and testing group effects from type i censored normal samples in experimental design

The maximum likelihood (ML) equations calculated from censored normal samples do not admit explicit solutions. A principle of modification is given and modified maximum likelihood (MML) equations, which admit explicit solutions, are defined. This approach makes it possible to tackle the hitherto unresolved problem of estimating and testing hypotheses about group-effects in one-way classification experimental designs based on Type I censored normal samples. The MML estimators of group-effects are obtained as explicit functions of sample observations and shown to be asymptotically identical with the ML estimators and hence BAN (best asymptotic normal) estimators. A statistic t is defined to test a linear contrast of group-effects and shown to be asymptotically normally distributed. A numerical example is presented which illustrates the procedure.