| Summary: | Clements and colleagues [1] claim there is an extreme decline effect in studies published between 2009 and 2019 on the impacts of ocean acidification (OA) on fish behaviour, with the modelled average effect size declining from >5 in 2009 to 2010 to <0.5 after 2015. Here, I show that the extreme decline effect reported by Clements and colleagues is a statistical artifact caused by the way they corrected for zero values in percentage data, which was more common in the earliest experiments compared with later studies. Furthermore, selective choices for excluding or including data, along with data compilation errors and missing studies with strong effects, weakened the effect sizes reported for papers after 2010, further exacerbating the decline effect reported by Clements and colleagues. When the data is reanalysed using appropriate corrections for zeros in percentage and proportional data and using a complete, corrected, and properly screened data set, the extreme decline effect reported by Clements and colleagues no longer exists (Fig 1A and 1B). Instead, there is a more gentle and consistent decline in effect size magnitude through time, from a modelled average <3 in 2009 to 2010 (Fig 1C) and remaining well above zero in 2018 to 2019 (Fig 1D).
|
|---|