Generalizing Research Findings for Enhanced Reproducibility: A Translational Medicine Case Study
13 Pages Posted: 12 Sep 2017 Last revised: 28 Mar 2019
Date Written: September 10, 2017
Reproducibility of research findings is a much-debated issue. A ramification of this discussion has led to statistical concerns for the interpretation of p-values. A related approach is focused on assessing false discovery rates (FDR). In this paper, we take a different perspective and discuss how to present findings derived from empirical studies that are claimed to be reproducible. This different perspective can help clarify the conflicting views expressed in the debate on reproducible research. Using an example from translational medicine, we demonstrate an approach for presenting research results by generalization of findings. Specifically, the methodology presented here maps research findings with a conceptual boundary of meaning (BOM) of alternative representations of the findings. A BOM delineates a demarcation line between alternative representations of research findings that share equivalence of meaning and alternatives, that appear similar, but have difference meanings not supported by the study. We suggest that it is the responsibility of researchers to present the BOM of their study. This mapping can be statistically evaluated by tests of hypotheses controlling for sign type (S-type errors) determined by the study design. Such errors apply to statements about the direction of an effect and are different from the traditional Ho or H1 statements applied to one sided or two sided tests of hypothesis. The example is of colon cancer detection with a composite polymeric system designed to target malignant regions with dual machinery. The research outcomes, determining affinity to colon cancer cell lines and malignant colon tissues from the rat model, were derived by statistically designed experiments. We suggest that in scientific publications, details on the study design experimental set up for repeatability purposes should be available only as supplementary material. Our main message is that reproducibility is more important than repeatability and that to inform about reproducibility characteristics, findings should be generalized.
Keywords: Reproducible research, generalizability of findings, information quality, boundary of meaning
Suggested Citation: Suggested Citation