It is pretty clear that the larger the heterogeneity, the harder it is to defend choosing a fixed effect model, as different studies cannot be assumed to provide estimates of a common, true effect.
Even within narrow clades of closely related genera and species, lack of absolute values from FE results means published stress and strain values cannot be used to evaluate relative performance of models across different studies (e.g. models of felid species in McHenry et al., 2007 versus those in Slater and Van Valkenburgh, 2009) [18], [22].
Subjective variables have been widely used in mathematical models in different studies such as market analysis (Nasrabadi et al. 2013), decision making (Hosseini and Tarokh 2013) and service quality evaluation (Soteriou and Stavrinides 1997).
We could not compare the forecast accuracy of models from different studies due to the lack of common measures and the lack of scale-independent measures.
Similarly, Schmid et al. (2006) found that efficiencies of the same model in different study areas can range from 0.583 (indicating very good model performance) to −0.911 (indicating bias).
We showed that each model has different study questions, assumptions, risk sets, and rate ratio interpretation, and so inferences should consider the appropriateness of the model for the RCT.
Bias in gene model identification among different studies might render a comparison of expansion or contraction challenging.
The random effects model assumes that different studies have substantial diversity and assesses both within-study sampling error and between-study variation [ 38].
The current study selects input model parameters from different studies as no single source could be identified for all of the necessary data.
In our review we also incorporated negative results, e.g., risk factors that were excluded from the final multivariate model in the different studies suggestive of lack of association.
