Do not log‐transform count data

By R. O’Hara, D. Kotze. 2010

1,282 citations recorded by the source corpus.

1. Ecological count data (e.g. number of individuals or species) are often log‐transformed to satisfy parametric test assumptions.

Subfields: Environmental Engineering, Nature and Landscape Conservation, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/b562850247de51ffcde4004818bd00d915817650

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Environmental Engineering

In Nature and Landscape Conservation

In Statistics and Probability

Citations in the corpus

Island, archipelago and taxon effects: mixed models as a means of dealing with the imperfect design of nature's experiments (2012, background): “However, we refer the reader to a recent study demonstrating that when faced with count data of this sort, a negative binomial or quasi-Poisson error structure often performs better than ln(n 1) transformation (O’Hara and Kotze 2010). Our fi xed eff ects were the area of the island (km 2 ), distan”
What you Feel, Is what you like Influence of Message Appeals on Customer Engagement on Instagram (2020, background): “Visual product centrality focuses directly on the features or benefits of the product and can be considered more of a hard sell than a soft sell (Okazaki, Mueller, and Taylor 2010).”
Multiple plant traits shape the genetic basis of herbivore community assembly (2015, methodology): “GLMs were appropriate because they account for response variables with non-normal distributions and heteroscedasticity that were not improved by transformations (O’Hara & Kotze 2010).”
Can You Gig It? An Empirical Examination of the Gig Economy and Entrepreneurial Activity (2018, positive): “…Tenreyro 2011), as in our sample, and to provide more reliable estimates, because a log-OLS specification in count data can produce severely biased estimates (O’Hara and Kotze 2010), particularly under conditions of heteroscedasticity, as a result of Jensen’s inequality (Silva and Tenreyro 2006).”
Modeling zero-inflated count data with glmmTMB (2017, positive): “These counts are often analyzed using generalized linear models (GLMs) and their extensions (O’Hara & Kotze, 2010; Wilson & Grenfell, 1997).”
What does a zero mean? Understanding false, random and structural zeros in ecology (2019, background): “Log transformation should be avoided because it does not guarantee that the key distributional assumptions (normality and homoscedasticity, among others) are satisfied and it fails to improve the distribution (O'Hara & Kotze, 2010).”