Experimental data and design, and the role of statistics

Abstract 

Data from experiments or observations come in many forms and may be categorized as either qualitative or quantitative. One requirement of a data set is to summarize it by expressing two pieces of information: a single value that ‘summarizes’ the data set – a measure of central tendency – and a notion of the variability of the data set. Methods to express these two descriptors are described, exemplified by definitions of the mean and median, standard deviation and standard error of the mean, as well as concepts such as confidence intervals and coefficient of variation. An experiment often involves introducing an intervention, or assessing the effects of an intervention on a data set. Hypothesis testing is a concept that gives information about the statistical significance of such an intervention in changing the values of a data set. Planning an experiment is an integral part of the scientific method. Experiments may be either observational studies (prospective or retrospective) or interventional. In each case bias must be avoided either when choosing the particular samples and their number that make up the data set, or in analysing the data and methods to minimize confounding variables. Finally, the principles of evidence-based medicine are introduced.

Keywords: data sets, evidence-based medicine, experimental design, hypothesis testing, sampling, statistics