Biomedical subjects
Publications and source records attributed to A Donner.
In this note we study, by simulation, small sample performance in terms of size of ten procedures for testing the homogeneity of odds ratios in K 2 x 2 contingency tables. These ten statistics are derived for 'large-stratum' settings. Our study concerns the behaviour of these statistics for 'small-stratum' settings.
Sample size requirements are provided for designs of studies in which clusters are randomized within each of several strata, where cluster size itself may be a stratifying factor. The approach generalizes a formula derived by Woolson et al., which provides sample size requirements for the Cochran-Mantel-Haenszel statistic. Issues of data analysis are also discussed.
We propose a new procedure for constructing a confidence interval about the kappa statistic in the case of two raters and a dichotomous outcome. The procedure is based on a chi-square goodness-of-fit test as applied to a model frequently used for clustered binary data. The procedure provides coverage levels that are accurate in samples of smaller size than those required for other procedures. The procedure also has use for significance-testing and the planning of corresponding sample size requirements.
There is recent interest in classification procedures intended for use only when the response is ordinal. Ordinal response, however, is evident in the parameters estimated by either multinomial logistic or normal discriminant analyses, both of which classify either ordinal or non-ordinal responses. Further, there may be harm in applying ordinal models inappropriately and ample opportunity to assume mistakenly ordinality in real data sets. Therefore, it becomes important to ascertain whether there is benefit obtained in the appropriate application of ordinal models. This paper presents the results of a simulation study designed to compare classification accuracy of various models. We show that ordinal models classify less accurately than the multinomial logistic and normal discriminant procedures under a variety of circumstances. Until further studies become available, we presently conclude that ordinal models confer no advantage when the main purpose of the analysis is classification.
McNemar's one degree of freedom chi-square test for the equality of proportions appears frequently in the analysis of pairs of matched, binary outcome data (Y1i, Y2i). An assumption underlying this test is that the responses from pair to pair are mutually independent. In certain applications, however, the pairs may represent repeated measurements on the same experimental unit, and hence this assumption is violated. In this paper we suggest an adjustment to the McNemar test to account for the repeated measures clustering effect and we report on a Monte Carlo simulation that evaluates the effectiveness of this approach.
There has been considerable methodologic research over the past 15 years on inference procedures for familial correlations. This paper reviews the various methods used to estimate sibling and parent-offspring correlations in family studies published over a recent 2-year period. The advantages and disadvantages of the various methodologic approaches are discussed, and practical recommendations given. One recommendation is for more frequent use of interval estimation. Computational formulas for selected approaches are presented in the appendices.
A unified non-iterative approach to point and interval estimation of interclass and intraclass correlations is presented in the context of family studies where there may be more than one individual in each of two classes. The procedure involves a generalization of the Pearson product-moment correlation coefficient, where one permits different weights for the pairs of scores. Unlike the maximum likelihood approach, these estimators are not derived under the assumption of a particular parametric form nor do they require an iterative solution. The asymptotic distributions of the generalized product-moment estimator and of the maximum likelihood estimator are derived under the assumption of normality. Also, several methods for constructing confidence intervals about the interclass correlation parameter are outlined, and the effectiveness of these methods is evaluated by Monte Carlo simulation. Although the focus of this paper is on the analysis of familial data, the methods discussed are applicable to more general situations, including the assessment of correlations between any two variables where each variable is replicated a different number of times for each sample unit.
Application of standard statistical procedures to site-specific data in periodontal research is invalid unless site-to-site dependencies are accounted for. In this paper, we present the adjustments required for valid application of matched pair procedures, including the paired t-test and McNemar's chi 2 test for correlated proportions. Examples are given involving data arising from: (1) the comparison of pre- and post-treatment clinical measurements; (ii) split-mouth protocols.
The McNemar chi-square statistic for testing the equality of proportions in matched pairs of binary data is expressed as a function of the Pearson chi-square statistic and a kappa statistic for measuring interrater agreement. Implications for data analysis and for sample size determination are discussed. The result is generalized to provide a similar relationship between Cochran's Q-statistic and the Pearson chi-square statistic for the comparison of proportions in matches of size m.
A methodological review is presented of 16 non-therapeutic intervention trials published over the last decade which have randomized intact clusters rather than individuals in treatment groups. Each of the trials was surveyed as to the information supplied on six methodological criteria. Although there is increasing recognition of the methodological issues associated with cluster randomization, many investigators are still not aware of the impact of this design on sample size requirements and analysis considerations. Investigators are urged to publish the cluster-specific event rates observed in their trials as a guide for the planning of future studies.
The asymptotic and finite-sample properties of several recent estimators of interclass correlation are compared to more traditional estimators in the case of a variable number of siblings per family. It is shown that Karlin's family-weighted pairwise estimator (Karlin, Cameron, and Williams, 1981, Proceedings of the National Academy of Science 78, 2664-2668) is virtually equivalent to the ensemble estimator (Rosner, Donner, and Hennekens, 1977, Applied Statistics 26, 179-187), thus suggesting an estimator of the former's asymptotic variance. Further, an estimator proposed by Srivastava (1984, Biometrika 71, 177-185) is shown to be identical to the modified sib-mean estimator (Konishi, 1982, Annals of the Institute of Statistical Mathematics 34, 505-515) when the sib-sib correlation is estimated by the method of unweighted group means. Although the estimator due to Srivastava has smaller asymptotic variance than the other two, the gain in efficiency is slight, for familial data, both asymptotically and in finite samples.
A 64-year-old female patient, whose extensive bilateral leg ulcers had been treated by daily dressings with diachylon ointment for more than a year, developed lead intoxication associated with general weakness, loss of weight, anemia, hypotension, and neuropathy. The intoxication was proved by the increased levels of lead in blood (threefold) and urine (tenfold). Treatment with dimercaptopropane sulphonic acid (Dimaval) resulted in a considerable emptying of the lead deposits, and the symptoms of poisoning subsided. It is discussed if lead-containing ointments are still justified in modern dermatology.
We develop two estimators of a common odds ratio psi for designs in which the investigator randomly assigns each of two clusters to interventions within strata. The estimators rely on an empirical adjustment for clustering to provide improved estimators of psi relative to the standard Woolf and Mantel-Haenszel estimators, respectively. The results of a simulation study show that the suggested adjustment improves the accuracy of both of these well-known estimators under conditions likely to arise in practice. We find the clustered Woolf estimator as particularly effective in terms of mean squared error reduction. We also discuss interval estimation.
We conducted a simulation study to determine the performance of nine procedures for testing the homogeneity of odds ratios in K 2 x 2 contingency tables. We recommend Tarone's approximate score test, based on the Mantel-Haenszel estimator of the common odds ratio, for use in practice. We also recommend a non-iterative statistic developed by Gart and based on the modified Woolf estimator of the common odds ratio for very large samples in balanced or mildly unbalanced designs. We base our recommendation of a statistic on its performance in terms of size and power in comparison with the other statistics considered.
Clinical studies in psychiatry frequently involve the analysis of matched pairs, in which each patient is matched with his or her psychiatric clinician. If the response variable of interest is dichotomous, then McNemar's chi 2 test for correlated proportions may be applied to the resulting data. However, this test assumes that the set of matched pairs are statistically independent, an assumption that is violated if some clinicians are matched with more than one patient. In this article we present alternatives to McNemar's test for dealing with this problem.
26 NPC patients and 47 TC patients were investigated clinically, and in the sera of these patients EBV-specific antibodies were determined. No relapse occurred in 11 patients with NPC and in 17 patients with TC during the period of 4 years. In addition to the determination of antibodies we looked for EBNA in 3 undifferentiated NPC and 4 undifferentiated TC, and for EBV-DNA in 2 of these NPC and one of these TC. The 3 NPC and 4 TC tested for EBNA showed this antigen in the cells. 2 NPC and 1 TC showed EBV-DNA with in situ hybridization technique. In the blot hybridization according to Southern the Eco R 1 fragment of both NPC was identical with the Eco R 1 fragment of the positive control of P 3 HR 1 cells. The 15 NPC-patients with tumour (4 with primary NPC, 11 with relapse) in the observation period showed higher IgA-VCA titres than the NPC patients free of relapse during this time. Subsequent to therapy, IgG-VCA titres decreased. Most of the NPC-patients with tumour during the time of testing mentioned before had IgA-VCA once or several times. Most of half of these patients had IgA-EA, but there was no patient without relapse. IgA-VCA increase and the presence of IgA-EA may be the first sign of a relapse. Thus, intensive research into this phenomenon will be required. No remarkable difference could be found in EBV-specific antibody patterns between TC patients without relapse and TC patients with tumour progression or relapse.
Donner and Banting (1988) presented a method of adjusting the standard chi-square test for the comparison of proportions when this procedure is applied to multiple-site dental data. The adjusted procedure allows for a comparison of two or more groups of subjects with respect to the prevalence over all sites of a specified characteristic, while taking site-to-site dependencies into account. In this paper, similar adjustments are presented for two other frequently used chi-square procedures, the chi-square test for linear trend and the Mantel-Haenszel test. Examples are given.
Ophthalmologic studies often compare several groups of subjects for the presence or absence of some ocular finding, where each subject may contribute two eyes to the analysis, the values from the two eyes being highly correlated. Rosner (1982, Biometrics 38, 105-114) and Dallal (1988, Biometrics 44, 253-257) proposed procedures for testing whether the proportion of affected eyes is the same among the different groups, while accounting for the intrasubject correlation. In this paper we propose an alternative approach, based on a simple adjustment of the standard Pearson chi-square test for the equality of proportions. The suggested approach utilizes information on subjects who supply only one eye to the analysis, and readily generalizes to studies in which more than two units of analysis are provided by each subject.