Fig. 3
From: Raiders of the lost HARK: a reproducible inference framework for big data science
A scheme for HARK-solid knowledge inference and model ensemble update. The orange compartments (real-world evidence and problem/research objective) represent the motivating forces for organizing a research endeavor. The gray compartments (hypotheses and study design) represent the intellectual effort to organize that research. The green compartments (available data/experiment sources, data sample and other data/experiments) represent the data one can use in pursuit of the inferences one seeks in regard to hypotheses or the construction of new models relevant to the real world. The purple components represent inference (new models) and knowledge bases (existing models). The magenta compartments (internal and external validation) authenticate inferences made by analyzing data with new or existing models: internal validation of inferences comes from study-generated models and data, while external validation comes from other data and models. The azure lines represent model generated actions that could lead to HARK. For the examples we have presented, this HARK is helpful for advancing science rather than generating non-reproducible results
