The validity of the chi-squared test when expected frequencies are small:A list of recent research references

A bibliography is given containing about 60 recent (chiefly post-1968) items on the validity of the chi-squared test for categorical data when the expected frequencies are small.     

The stability of several measures of association in small contingency tables

Measures of association are often used to describe the relationship between row and column variables in two—dimensional contingency tables. It is not uncommon in biomedical research to categorize continuous variables to obtain a two—dimensional table. In these situations it is desirable that the measure of association not be too sensitive to changes in the number of categories or to the choice of cut points. To accomplish this objective we attempt to find a measure of association that closely approximates the corresponding measure of association for the underlying distribution.Measures that are close to the underlying measure for various table sizes andcutpoints are called stable measures.     

Case–control studies with complex sampling

The use of complex sampling designs in population-based case–control studies is becoming more common, particularly for sampling the control population. This is prompted by all the usual cost and logistical benefits that are conferred by multistage sampling. Complex sampling has often been ignored in analysis but, with the advent of packages like SUDAAN, survey-weighted analyses that take account of the sample design can be carried out routinely. This paper explores this approach and more efficient alternatives, which can also be implemented by using readily available software.     

The effect of category choice on the odds ratio and several measures of association in case-control studies

In many case-control studies the risk factors are categorized in order to clarify the analysis and presentation of the data. However, inconsistent categorization of continuous risk factors may make interpretation difficult. This paper attempts to evaluate the effect of the categorization procedure on the odds ratio and several measures of association. Often the risk factor is dichotomized and the data linking the risk factor and the disease is presented in a 2 x 2 table. We show that the odds ratio obtained from the 2x2 table is usually considerably larger than the comparable statistic that would have been obtained had a large number of outpoints been used. Also, if 2 x 2, 2 x 3, or 2 x 4 tables are obtained by using a few outpoints on the risk factor, the measures of association for these tables are usually greater than the measure that would have been obtained had a large number of cntpoints been used. We propose an odds ratio measure that more closely approximates the odds ratio between the continuous risk factor and disease. A corresponding measure of association is also proposed for 2 x 2, 2x3, and 2x4 tables.     

Covariate Order Tests for Covariate Effect

A new approach for constructing tests for association between a random right censored life time variable and a covariate is proposed. The basic idea is to first arrange the observations in increasing order of the covariate and then base the test on a certain point process defined by the observation times. Tests constructed by this approach are robust against outliers in the covariate values or misspecification of the covariate scale since they only use the ordering of the covariate. Of particular interest is a test based on the Anderson-Darling statistic. This test has good power properties both against monotonic and nonmonotonic dependencies between the covariate and the life time variable.     

Bias due to Ignoring the Sample Design in Case–Control Studies

Case–control studies allow efficient estimation of the associations of covariates with a binary response in settings where the probability of a positive response is small. It is well known that covariate–response associations can be consistently estimated using a logistic model by acting as if the case–control (retrospective) data were prospective, and that this result does not hold for other binary regression models. However, in practice an investigator may be interested in fitting a non–logistic link binary regression model and this paper examines the magnitude of the bias resulting from ignoring the case–control sample design with such models. The paper presents an approximation to the magnitude of this bias in terms of the sampling rates of cases and controls, as well as simulation results that show that the bias can be substantial.     

A bayesian analysis of the multinomial model for a dichotomous response with nonrespondents

This paper studies the mu1tinomial model 2x2 contingency table data with some cell counts missing .Various hypotheses of interest including row-column independence are tested by using Bayes factors which represent the ratio of the posterior odds to the prior odds for the null hypothesis. The Dirichlet-Beta family of prior distributions is considered for the multinomial parameters cond itional on the complement of the null hypothesis. The Bayes factor for the incomplete data is a mixture of the Bayes factors for different possibilities for the full data.     

On the inferential use of predictive sample reuse

The predicitive sample reuse (PSR) data analysis technique proposed by Geisser and Eddy (1979) is applied to the analysis of categorical data. This application yiclds a new approach which has a number of advantages over classical methods for analysis of such data. A PSR technique for comparing linear or nonlinar regressino relationships, for two or more populations, and a PSR ailernative to certain nonparaemetri statistical tesis are also proposed.     

Tests for random time effects and spatial error correlation in panel regression models

We consider a panel model with spatial autocorrelation and heterogeneity across time. Various Lagrange multiplier and likelihood ratio test statistics are developed for testing time effects and spatial effects, jointly, marginally or conditionally. Limiting null distributions of the tests are derived. Size and power performances of the proposed tests are compared by a Monte-Carlo experiment.     

Non‐Parametric Change‐Point Tests for Long‐Range Dependent Data

Abstract. We propose a non‐parametric change‐point test for long‐range dependent data, which is based on the Wilcoxon two‐sample test. We derive the asymptotic distribution of the test statistic under the null hypothesis that no change occurred. In a simulation study, we compare the power of our test with the power of a test which is based on differences of means. The results of the simulation study show that in the case of Gaussian data, our test has only slightly smaller power minus.3pt than the ‘difference‐of‐means’ test. For heavy‐tailed data, our test outperforms the ‘difference‐of‐means’ test.     

Maximum Likelihood Estimation for Multinomial‐Poisson Models: A Generalization of Birch's Numerical Invariance Results

Abstract. This study gives a generalization of Birch's log‐linear model numerical invariance result. The generalization is given in the form of a sufficient condition for numerical invariance that is simple to verify in practice and is applicable for a much broader class of models than log‐linear models. Unlike Birch's log‐linear result, the generalization herein does not rely on any relationship between sufficient statistics and maximum likelihood estimates. Indeed the generalization does not rely on the existence of a reduced set of sufficient statistics. Instead, the concept of homogeneity takes centre stage. Several examples illustrate the utility of non‐log‐linear models, the invariance (and non‐invariance) of fitted values, and the invariance (and non‐invariance) of certain approximating distributions.     

Likelihood Ratio Type Two‐Sample Tests for Current Status Data

Abstract. We introduce fully non‐parametric two‐sample tests for testing the null hypothesis that the samples come from the same distribution if the values are only indirectly given via current status censoring. The tests are based on the likelihood ratio principle and allow the observation distributions to be different for the two samples, in contrast with earlier proposals for this situation. A bootstrap method is given for determining critical values and asymptotic theory is developed. A simulation study, using Weibull distributions, is presented to compare the power behaviour of the tests with the power of other non‐parametric tests in this situation.     

On the Hypothesis of No Interaction on a Linear Scale in Contingency Tables

The purpose of this paper is to review briefly the three main formulations of no Interaction hypotheses in contingency tables and to consider the formulation on a linear scale in some detail.More specifically we (i) present a situation in 2×2 tables where such a formulation may be more appropriate than others, (ii) study the geometry for this problem, (iii) give contrast-type or parametric ANOVA type formulations in the general n-dimensional tables, (iv) discuss estimation and testing procedures and (v) consider collapsibility of contingency tables in relation to the hypotheses of no interaction on a linear scale.     

The minimum discrimination information approach in analyzing categorical data

A brief review of the minimum discrimination information (MDI) approach in analyzing categorical data is presented in a question -answer format, An example is given to bring out situations in which the MDI approach is more useful. No new results are proved.     

A bayesian method for inferring the degree fo dependence for a positively quadrant dependent distribution

A Bayesian method for making inferences on the degree of dependence for a positively quadrant dependent distribution is developed. This method is based on a parametric model, where the parameter measures the degree of association. The parametric model is a natural simplification of the original one: it is a mixture of the two extreme cases corresponding to independence and positive extremal dependence.     

A modified chi-square test of independence against a class of ordered alternatives in an r x c contingency table

A modified chi-square test statistic is constructed for testing the hypothesis of independence in a two-way contingency table against a class of ordered alternatives defined in terms of pooled cross-product ratios. The test procedure can also be used to test for positive quadrant dependence in a two-way contingency table. The asymptotic distribution of the test statistic under the null hypothesis is obtained. Some power comparisons with known test procedures are given. A numerical example is given to illustrate the use of this test.     

Tests of odds ratio homogeneity with improved power in sparse fourfold tables

Some new tests of odds ratio homogeneity for fourfold tables are compared with the mixture model score test in the sparse-data case (many tables, small margins per table). Based on general empirical Bayes inequalities, the new tests have competitive power for 1:R matched designs, and superior power for more balanced designs.     

Correction of the p-Value after Multiple Tests in a Cox Proportional Hazard Model

We consider a situation which is common in epidemiology, in which several transformations of an explanatory variable are tried in a Cox model and the most significant test is retained. The p-value should then be corrected to take account of the multiplicity of tests. Bonferroni method is often too conservative because the tests may be highly positively correlated. We propose an asymptotically exact correction of the p-value. The method uses the fact that the tests are asymptotically normal to compute numerically the distribution of the maximum of several tests. Counting processes theory is used to derive estimators of the correlations between tests. The method is illustrated by a simulation and an analysis of the relation between concentration of aluminum in drinking water and risk of dementia.     

Conditional tests in a competing risks model

Testing for equality of competing risks based on their cumulative incidence functions (CIFs) or their cause specific hazard rates (CSHRs) has been considered by many authors. The finite sample distributions of the existing test statistics are in general complicated and the use of their asymptotic distributions can lead to conservative tests. In this paper we show how to perform some of these tests using the conditional distributions of their corresponding test statistics instead (conditional on the observed data). The resulting conditional tests are initially developed for the case of k = 2 and are then extended to k > 2 by performing a sequence of two sample tests and by combining several risks into one. A simulation study to compare the powers of several tests based on their conditional and asymptotic distributions shows that using conditional tests leads to a gain in power. A real life example is also discussed to show how to implement such conditional tests.