Relevance of statistical signif icance in medical research

Abstract

In order to answer a research question, observations in the form of data are obtained from a sample of individuals or subjects under study. The sample so obtained should be random in nature. There are random sampling methods to obtain samples that adequately represent a cross-section of the entire population. Samples of observations are used to draw valid conclusions. In this approach, the quantification of uncertainties that occur due to the process of sampling is the subject matter of evaluating statistical significance. These are different from the uncertainties related to measurement error in either the instrument or observers occurring due to non-sampling errors, which are not discussed here. Statistical significance is closely related to confidence statements of occurrence of an event. One can never be 100% confident in drawing conclusions. A threshold of 95% confidence indicates an uncertainty of 5% which could result in errors. The statistical inference in hypothesis testing deals with the probability of those errors.

Keywords

• sample
• standard deviation
• confidence interval
• null hypothesis