Dataset Information

To test or not to test: Preliminary assessment of normality when comparing two independent samples.

ABSTRACT:

Background

Student's two-sample t test is generally used for comparing the means of two independent samples, for example, two treatment arms. Under the null hypothesis, the t test assumes that the two samples arise from the same normally distributed population with unknown variance. Adequate control of the Type I error requires that the normality assumption holds, which is often examined by means of a preliminary Shapiro-Wilk test. The following two-stage procedure is widely accepted: If the preliminary test for normality is not significant, the t test is used; if the preliminary test rejects the null hypothesis of normality, a nonparametric test is applied in the main analysis.

Methods

Equally sized samples were drawn from exponential, uniform, and normal distributions. The two-sample t test was conducted if either both samples (Strategy I) or the collapsed set of residuals from both samples (Strategy II) had passed the preliminary Shapiro-Wilk test for normality; otherwise, Mann-Whitney's U test was conducted. By simulation, we separately estimated the conditional Type I error probabilities for the parametric and nonparametric part of the two-stage procedure. Finally, we assessed the overall Type I error rate and the power of the two-stage procedure as a whole.

Results

Preliminary testing for normality seriously altered the conditional Type I error rates of the subsequent main analysis for both parametric and nonparametric tests. We discuss possible explanations for the observed results, the most important one being the selection mechanism due to the preliminary test. Interestingly, the overall Type I error rate and power of the entire two-stage procedure remained within acceptable limits.

Conclusion

The two-stage procedure might be considered incorrect from a formal perspective; nevertheless, in the investigated examples, this procedure seemed to satisfactorily maintain the nominal significance level and had acceptable power properties.

SUBMITTER: Rochon J

PROVIDER: S-EPMC3444333 | biostudies-literature | 2012 Jun

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

To test or not to test: Preliminary assessment of normality when comparing two independent samples.

Rochon Justine J Gondan Matthias M Kieser Meinhard M

BMC medical research methodology 20120619

<h4>Background</h4>Student's two-sample t test is generally used for comparing the means of two independent samples, for example, two treatment arms. Under the null hypothesis, the t test assumes that the two samples arise from the same normally distributed population with unknown variance. Adequate control of the Type I error requires that the normality assumption holds, which is often examined by means of a preliminary Shapiro-Wilk test. The following two-stage procedure is widely accepted: If ...[more]

PMID: 22712852

Similar Datasets

Project description:BackgroundInternalizing (anxiety and mood) disorders (INTD) commonly co-occur (are "comorbid") with alcohol use disorder (AUD). The literature suggests that excessive alcohol use aimed at coping with INTD symptoms is, at best, a partial explanation for the high comorbidity rates observed. We hypothesized that individuals with INTD experience greater susceptibility to developing AUD symptoms due to the partially shared neurobiological dysfunctions underlying both conditions. We probe this hypothesis by testing the prediction that, after accounting for the volume of alcohol intake, individuals with INTD experience higher levels of alcohol-related symptoms.MethodsData from the National Epidemiological Survey on Alcohol-Related Conditions (NESARC) Wave 3 were used for the primary analyses, and NESARC Wave 1 data were used for independent replication analyses. Individuals who reported any alcohol use in the prior year were categorized as: (1) never having had an INTD diagnosis ("INTD-Never"); (2) having a remitted INTD diagnosis only ("INTD-Remitted"); or (3) having current INTD diagnosis ("INTD-Current"). Between-group contrasts of alcohol-related symptoms controlled for total alcohol intake (past year), drinking pattern (e.g., binging) and variables previously shown to mark exaggerated AUD symptoms relative to drinking amount (e.g., SES, gender, and family history).ResultsWith all covariates in the model, individuals in the INTD-Current group and the INTD-Remitted group reported significantly greater alcohol-related symptoms than those in the INTD-Never group but did not themselves differ in level of alcohol-related symptoms. These results were replicated in the NESARC 1 dataset.ConclusionsIndividuals with INTD experience more alcohol-related symptoms than those who drink at the same level. While considering other explanations, we argue that this "harm paradox" is best explained by the view that INTD confers a neurobiologically mediated susceptibility to the development of AUD symptoms.

Project description:The new generation of whole genome sequencing platforms offers great possibilities and challenges for dissecting the genetic basis of complex traits. With a very high number of sequence variants, a naïve multiple hypothesis threshold correction hinders the identification of reliable associations by the overreduction of statistical power. In this report, we examine 2 alternative approaches to improve the statistical power of a whole genome association study to detect reliable genetic associations. The approaches were tested using the Genetic Analysis Workshop 19 (GAW19) whole genome sequencing data. The first tested method estimates the real number of effective independent tests actually being performed in whole genome association project by the use of an extreme value distribution and a set of phenotype simulations. Given the familiar nature of the GAW19 data and the finite number of pedigree founders in the sample, the number of correlations between genotypes is greater than in a set of unrelated samples. Using our procedure, we estimate that the effective number represents only 15 % of the total number of independent tests performed. However, even using this corrected significance threshold, no genome-wide significant association could be detected for systolic and diastolic blood pressure traits. The second approach implements a biological relevance-driven hypothesis tested by exploiting prior computational predictions on the effect of nonsynonymous genetic variants detected in a whole genome sequencing association study. This guided testing approach was able to identify 2 promising single-nucleotide polymorphisms (SNPs), 1 for each trait, targeting biologically relevant genes that could help shed light on the genesis of the human hypertension. The first gene, PFH14, associated with systolic blood pressure, interacts directly with genes involved in calcium-channel formation and the second gene, MAP4, encodes a microtubule-associated protein and had already been detected by previous genome-wide association study experiments conducted in an Asian population. Our results highlight the necessity of the development of alternative approached to improve the efficiency on the detection of reasonable candidate associations in whole genome sequencing studies.

Project description:Previous studies have proposed a simple product-based estimator for calculating exposure-specific risks (ESR), but the methodology has not been rigorously evaluated. The goal of our study was to evaluate the existing methodology for calculating the ESR, propose an improved point estimator, and propose variance estimates that will allow the calculation of confidence intervals (CIs).We conducted a simulation study to test the performance of two estimators and their associated confidence intervals: 1) current (simple product-based estimator) and 2) proposed revision (revised product-based estimator). The first method for ESR estimation was based on multiplying a relative risk (RR) of disease given a certain exposure by an overall risk of disease. The second method, which is proposed in this paper, was based on estimates of the risk of disease in the unexposed. We then multiply the updated risk by the RR to get the revised product-based estimator. A log-based variance was calculated for both estimators. Also, a binomial-based variance was calculated for the revised product-based estimator. 95% CIs were calculated based on these variance estimates. Accuracy of point estimators was evaluated by comparing observed relative bias (percent deviation from the true estimate). Interval estimators were evaluated by coverage probabilities and expected length of the 95% CI, given coverage. We evaluated these estimators across a wide range of exposure probabilities, disease probabilities, relative risks, and sample sizes.We observed more bias and lower coverage probability when using the existing methodology. The revised product-based point estimator exhibited little observed relative bias (max: 4.0%) compared to the simple product-based estimator (max: 93.9%). Because the simple product-based estimator was biased, 95% CIs around this estimate exhibited small coverage probabilities. The 95% CI around the revised product-based estimator from the log-based variance provided better coverage in most situations.The currently accepted simple product-based method was only a reasonable approach when the exposure probability is small (< 0.05) and the RR is ? 3.0. The revised product-based estimator provides much improved accuracy.

Dataset Information

To test or not to test: Preliminary assessment of normality when comparing two independent samples.

Background

Methods

Results

Conclusion

Publications

To test or not to test: Preliminary assessment of normality when comparing two independent samples.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets