Statistical relevance--relevant statistics, part I.

S tatistical analysis is an important tool in experimental research and is essential for the reliable interpretation of experimental results. It is essential that statistical design should be considered at the very beginning of a research project, not merely as an afterthought. For example if the sample size for an experiment only allows for an underpowered statistical analysis, then the interpretation of the experiment will have to be limited. An experiment cannot be reverse engineered to become more statically significant, although experiments can of course be repeated independently to account for biological variation (see section on technical versus biological replicates below). Statistical methods are tools applied to situations in which we encounter variability, noise and uncertainty. They help make more definitive scientific conclusions, and to make better use of available resources. In this new EMBO Journal statistics series, I will introduce key concepts and best practices. The text will be short and conceptual in style, while the supplement will provide examples demonstrating the introduced concepts. I will use the free statistic software R (R Core Team, 2015) to illustrate examples, and readers can try the code on their own data. In this first part, I will give some guidelines for initial study design and analysis of experiments. Subsequent columns will discuss specific statistical topics in more detail. Most of the issues touched upon in this first column are further discussed in the book of Ruxton and Colegrave (Ruxton & Colegrave, 2010), which includes many examples relevant to the analysis of experiments for biological researchers. Guidelines and terms for the design and analysis of experiments