Essentials of Biostatistics. 8. Basic philosophy of statistical tests, confidence intervals and sample size determination

Abstract

In empirical sciences such as medicine, an inference or a decision one way or the other is taken on the basis of samples in full realization that the sampling fluctuations are present. These fluctuations occur since the individuals included in different samples differ from each other. This intersample variability leads to uncertainties. Fortunately, statistical methods allow us to draw valid conclusions by assessing the expected magnitude of intersample variability on the basis of just one sample. The intersample variability can be assessed by studying, what is called, sampling distribution. This is dis-cussed in Section 8.1. This sampling distribu-tion is the basis for the statistical inference. The inference is basically of two types. We describe confidence intervals (CIs) in Section 8.2 and statistical significance in Section 8.3. For significance, we discuss only the concept in this Article. The actual methods will be presented in subsequent articles of this series. Sample size determination is intimately related to the concept of CI and of test of significance. We discuss some facets of this in Section 8.4.