Statistics in a Nutshell, 2nd Edition

Chapter 18. Research Design

Often, one of the responsibilities of a statistician is to design research studies. To do this well, you must be familiar with the different types of research designs, know their strengths and weakness, and be able to draw on this knowledge to design studies to examine different types of questions. You also need to be familiar with the customs and practices of your profession, such as what type of study is generally used for a particular type of data or to answer a particular type of question. Research design is a larger subject than can be covered in a single chapter, so this chapter can only introduce the major issues in designing research studies and discuss some of the most common types of designs. Typically, designing a study involves compromise between what the researcher would ideally like to do and what is feasible, and the choice and execution of a design should be guided by consideration of what is most important to the research question and the traditions and standard practices in the relevant field of study. We’d all love to conduct research that is both perfectly controlled (meaning the experimenter can manipulate or otherwise control all the factors relevant to the research) and perfectly naturalistic (meaning that whatever is being measured is being so in a realistic, natural environment). However, the characteristics of control and naturalism are often in competition with each other, and learning to judge how much to emphasize one versus ...