Alternative to Tukey test

ABSTRACT In order to search for an ideal test for multiple comparison procedures, this study aimed to develop two tests, similar to the Tukey and SNK tests, based on the distribution of the externally studentized amplitude. The test names are Tukey Midrange (TM) and SNK Midrange (SNKM). The tests were evaluated based on the experimentwise error rate and power, using Monte Carlo simulation. The results showed that the TM test could be an alternative to the Tukey test, since it presented superior performances in some simulated scenarios. On the other hand, the SNKM test performed less than the SNK test.

A New Approach to the Problem of Multiple Comparisons in the Genetic Dissection of Complex Traits

Saturated genetic marker maps are being used to map individual genes affecting quantitative traits. Controlling the “experimentwise” type-I error severely lowers power to detect segregating loci. For preliminary genome scans, we propose controlling the “false discovery rate,” that is, the expected proportion of true null hypotheses within the class of rejected null hypotheses. Examples are given based on a granddaughter design analysis of dairy cattle and simulated backcross populations. By controlling the false discovery rate, power to detect true effects is not dependent on the number of tests performed. If no detectable genes are segregating, controlling the false discovery rate is equivalent to controlling the experimentwise error rate. If quantitative loci are segregating in the population, statistical power is increased as compared to control of the experimentwise type-I error. The difference between the two criteria increases with the increase in the number of false null hypotheses. The false discovery rate can be controlled at the same level whether the complete genome or only part of it has been analyzed. Additional levels of contrasts, such as multiple traits or pedigrees, can be handled without the necessity of a proportional decrease in the critical test probability.

Is There a Relationship between Self Esteem and Drug Use? Methodological and Statistical Limitations of the Research

Although a causal connection between self-esteem and drug use might make intuitive sense, a critical evaluation of the research calls this relationship into question. The most fatal flaw in the “low self-esteem causes drug use” argument is the fact that only a very small proportion of the variance in drug use is associated with self-esteem across a variety of definitions of self-esteem. In addition, the literature is fraught with methodological and statistical problems that severely limit the conclusions that can be drawn. Methodological problems examined in the article include: measurement of self-esteem, measurement of drug use and abuse, inclusion of confounding variables, and tendency to infer causality from correlational data. Statistical problems explored are: differences between the results of studies employing multivariate and bivariate statistics, failure to report strength of association indices, inflated experimentwise error rate when conducting numerous statistical analyses, failure to collapse several highly correlated variables into fewer factors, tendency to misinterpret statistical data, and reporting insufficient statistical information to allow readers to draw their own conclusions. We conclude that the scientific evidence relating self-esteem and drug use is insufficient to justify making self-esteem enhancement the cornerstone of drug prevention efforts.

Simultaneous Multiple Comparison Procedures in Psychiatric Research

Methods are presented for using linear contrasts to make inferences about differences between the means of several populations on continuous dependent variables. These methods control the experimentwise error rate (the probability of committing one or more type 1 errors in the set of decisions made within the experiment) for linear contrasts which compare some sub-sets of populations with others. Appropriate methods are outlined for testing contrasts which have been planned (i.e., specified independently of the data on which they are tested) and defined post hoc (i.e., after an inspection of the data). We show how these methods can be adapted to the analysis of data from factorial analysis of variance research designs.