Exploratory approaches and RRR involve arbitrary choices such as consolidation of food
items into food groups, number of DPs to be retained, selection of response variables in
RRR, applied rotation method in factor analysis, and naming of the DPs [49]. While for the
hypothesis-based approaches the inclusion of the selected nutrients or foods largely depends
on their presence in the dietary recommendations utilized to define the score, they may also
depend on the investigator’s arbitrary selection of dietary components, weightings and
cutoff points of them. McCann et al. reported that the choice of food groups had only a
small impact on the observed associations between PCA-identified DPs and endometric
cancer [50]. In another study, we examined the effect of different combinations of response
variables on RRR-derived DPs and their relationship with AD risk, and the results showed a
“common” DP, with very similar food components, can be identified and its association with
AD risk was quite stable [48]. Although these two studies suggest that the obtained DPs are
independent of parameter selections involved, more studies are warranted to test the
reproducibility of the results.