Table 2.

Comparison of perspectives.

	Summary	Rationale	Notes	Relevant to discussions and debates about
With vs. without a mediation hypothesis				TE power False-positive psychology
	Statistical significance of the TE is irrelevant for testing mediation	A mediation hypothesis is required before testing for indirect effects The test of IE often has greater power than the test of the TE1, and requiring TE1 be significant ignores this, as well as the risk of opposing direct and indirect	The basis of both NHST and theory-driven research necessitate that tests are done carefully, and statistical significance is not “fished” for
	The TE should be significant before testing mediation	The basis of both NHST and theory-driven research necessitate that tests are done carefully, and statistical significance is not “fished” for by moving to a IE test after non-significance of TE1 test	Does not apply to exploratory research conducted in a manner that is in keeping with best practices
Specific Effects vs. Global Model				SEM vs. regression Network models
	Effects perspective focuses on individual parameters, with inferences based on their presence and/or strength	Often it is only necessary to show that an effect exists, generally in support of a given theoretical framework	Model-constraints that may bias parameters are avoided, but runs the risk of overfitting and poor replication rates due to larger standard errors
	Statistical models are the focus, with mediation effect(s) a subset of the global set of relationships	Effects do not occur in isolation, and are instead a part of a larger set of causes, effects, and boundary conditions	Higher precision and better replication if model assumptions hold, but biased parameter estimates if they do not. Interpretation of the effects is also conditional on the global model used
Significance testing vs. Effect sizes				Theory building vs. practical application
	Null hypothesis significance testing (NHST) is used as the criteria to evaluate an effect	Priority is to distinguish an effect from noise	Ignores gradation of effects and does not address proportion of outcome variance explained
	Effect sizes and confidence are used to evaluate an effect	Establishing presence is insufficient because uncertainty and magnitude affect both replication and relevance of a given effect	Meaningful effects may be ignored because they seem too small to matter in practice, but are nonetheless important when aggregated across many people or for theory building
Direct vs. Indirect Effects				Appropriateness of indirect effects
	A directness perspective focuses establishing direct causes	For some problems and situations one may want to know direct causes, intermediate steps leave room for interference	Has the advantage that one does not need to count on uncertain intermediate steps, but potentially ignores suppression effects due to other variables
	An indirectness perspective focuses on understanding the intermediate steps between IV and DV, between a cause and its effect	For almost any given cause and effect encountered there are intermediate effects, and explaining and understanding these effects provides more information about the original cause and effect mechanisms which may themselves be more relevant to the outcome	Provides a fuller picture of the relationship between X and Y, but at much greater risk of making incorrect inferences and explanations
Hypothesized vs. Alternative explanations				False-positive psychology Parameter sensitivity
	Focus is on confirmatory evidence and supporting a mediation hypothesis	Past research should guide future research, increasing speed and ideally efficiency of research	Restricted explanation search that runs the risk of attempting to support previous findings and neglecting more accurate alternative explanations
	Focus is on testing alternative hypotheses that conflict with a specific mediation hypothesis	Evidence in support of a mediation claim is stronger if it can be shown that plausible alternative explanations do not hold	Infinite number of alternative explanations, both reasonable and unreasonable, and often no clear stopping rule for what constitutes an adequate search