Priors & prejudice : using existing knowledge in social science research

Abstract

Researchers in the social sciences usually start their research with the formulation of research goals and questions, which, together with studying the existing literature, lead to the formulation of hypotheses. Next, data is collected using experiments or questionnaires and is subsequently analyzed. Finally, conclusions are drawn based on the research findings. Although this procedure is a logical one, more scientific knowledge can be gained by starting a research project with a sharper focus. In this dissertation a sharper focus is established by formulating and analyzing informative hypotheses. Informative hypotheses contain information that is known beforehand, i.e., they contain existing knowledge. Such hypotheses are usually formulated in words in the introduction of a scientific article, but they are hardly ever analyzed. In order to adequately analyze informative hypotheses stated in words, they need to be translated into statistical hypotheses. This can be done by placing (in)equality signs (, =) between the parameters of interest (e.g., means or regression coefficients). These so called inequality constrained hypotheses can be used in statistical analyses. This dissertation offers researchers in the social sciences instruments, tools and analysis strategies to help them get better answers to their research questions by using informative hypotheses. The following points are addressed: 1) Are researchers willing to use new research methods? 2) How can informative hypotheses be formulated? And what existing knowledge should be included? 3) How can informative hypotheses be analyzed using classical and Bayesian statistics? and 4) What are the advantages and disadvantages of these analysis techniques? Furthermore, an example concerning a study on therapy effect for girls exposed to a single rape is given. From this dissertation it can be concluded that usage of informative hypotheses is an elegant way of including existing knowledge in current research. Reasons for using informative hypotheses are: 1) Existing customs with concern to conducting research do not have to change much; 2) More attention is paid to scientific knowledge that was previously gained; 3) Existing knowledge is incorporated in a formal and transparent manner; and 4) Research questions get more accurate answers