Statistics in Nuclear Cardiology: So, What’s the Difference? The t-Test: Pitfalls and Options in Hypothesis Testing for Comparing Differences in Means

抄録

Decisions related to differences of the measures of central tendency of population parameters are an important part of clinical research. Choice of the appropriate statistical test is critical to avoiding errors when making those decisions. All statistical tests require that one or more assumptions be met. The t-test is one of the most widely used tools but is not appropriate when assumptions such as normality are not met, especially when small samples, <40, are used. Non-parametric tests, such as the Wilcoxon rank sum and others, offer effective alternatives when there are questions about meeting assumptions. When normality is in question, the Wilcoxon non-parametric tests offer substantially higher levels of power and an reliable alternative to the t-test.