1b. Title of research proposal:
Experts, their prior knowledge, and the issue of limited data
1c.Summary of research proposal
Researchers often have difficulties collecting enough data to obtain statistical power: target groups are small (e.g., children with severe burn injuries), hard to access (e.g., infants of drug-dependent mothers), or come along with prohibitive costs (e.g., measuring phonological difficulties of babies). Such obstacles to gathering data leads to a limited data set. To deal with this, researchers can simplify their hypotheses and run simplified statistical models. This strategy, however, is undesirable since it does not provide an answer to the intended research question. As I investigated in my VENI-project, researchers may overcome the issue of limited data by using Bayesian statistics, but only if background knowledge is available and can be translated into statistical ‘prior distributions’. However, background knowledge based on previous publications, as I used in my VENI-project, might not be available for limited data sets since such data usually remains unpublished. Therefore, given my background in statistics as well as developmental psychology, I propose to use experts’ opinions (researchers, clinicians and experts-by-experience) to provide the necessary information. Yet, the road one needs to take to incorporate expert knowledge is challenging: Methods have to be developed to assist experts in specifying statistical prior distributions; If the information from experts contradicts the information in the data, tools (and guidelines) are needed to indicate whether this ‘expert-data conflict’ is substantial; Also, methods are needed to deal with biased or disagreeing experts. In the current proposal I set out a program to overcome these obstacles through a cutting edge methodology taking the research in this area a major leap forward. In addition, since I have a large network of researchers in many different domains of the social sciences, I will use ‘real-life’ examples to provide a strong test and demonstration of the potential advantages and pitfalls of the proposed methodology.