| 要旨トップ | 目次 | 日本生態学会第66回全国大会 (2019年3月、神戸) 講演要旨
ESJ66 Abstract


一般講演(口頭発表) G01-08  (Oral presentation)

The (mis) assessment of ecological risk for nickel based on the conventional statistical approach in ecotoxicology: the importance of causal inference

*Kazutaka M Takeshita, Takehiko I Hayashi, Hiroyuki Yokomizo(NIES)

To illustrate the importance of causal inference in analyses relating to decision-making of management intervention, we assessed ecological risk of nickel based on the following two approaches: (1) a correlation analysis, quantile regression, which is a method conventionally used in ecotoxicology, and (2) a multiple regression model with covariates satisfying the back-door criterion for estimating the causal effect of nickel. To construct such multiple regression model, we developed the causal diagram comprised of the biological indicator (Ephemeroptera, Plecoptera, and Trichoptera [EPT] richness), quality of river water, and physical characteristics of river. We used a dataset of water quality and benthic invertebrates collected at 45 sites in 19 rivers. There was the negative association between EPT richness and nickel concentrations in the 90th quantile regression model, however, the causal effect of nickel concentrations on EPT richness was not confirmed in the multiple regression model. There was a significant causal effect of flow velocity on EPT richness. Both EPT richness and nickel concentration differed in each river, therefore, the spurious association resulted from a confounding factor, a variable of "river". Management interventions in nickel concentration determined by ecological risk assessment based on correlation analysis may not lead to an expected outcome.


日本生態学会