Visual heuristics for marginal effects plots:

Research & Politics

Thomas B. Pepinsky¹ 

关键词: Interaction terms;    marginal effects plots;    conditional hypotheses;    data visualization;   

DOI  :  10.1177/2053168018756668

学科分类：社会科学、人文和艺术（综合）

来源: Sage Journals

【 摘 要 】

Common visual heuristics used to interpret marginal effects plots are susceptible to Type-1 error. This susceptibility varies as a function of (a) sample size, (b) stochastic error in the true data generating process, and (c) the relative size of the main effects of the causal variable versus the moderator. I discuss simple alternatives to these standard visual heuristics that may improve inference and do not depend on regression parameters.

【 授权许可】

CC BY-NC-ND