The preprint by Jardim et al. makes important advances in the assessment of missing trait data. They analyzed how the simulated target and auxiliary functional traits imputed in different sorts of missing data would influence the descriptive statistics, model parameters, and phylogenetic signal estimation from these databases. They simulated coalescent trees and missing data (missing completely at random, missing at random but phylogenetically structured, and missing at random but correlated with another variable) and found that the structure of the missing data, the evolutionary model used to simulate the phylogeny and the percentage of missing data were important factors determining estimation errors. The manuscript is well-written and makes a novel and sound contribution to the ecological literature. Trait data is rarely available for entire communities of species and most trait databases use imputation methods. In spite of the common use, previous studies have not evaluated the impact of data imputation in common metrics of trait distribution and phylogenetic signal. This paper shows that missing trait data and data imputation can create biases in common ecological and evolutionary metrics, and suggest ways to minimize the problem when only incomplete data are available. Our suggestions are outlined below.
We are not sure whether authors challenged the shortfall and the consequences of using imputed databases. If they only challenged the shortfall (and not the consequences), the title could be: “Challenging the Raunkiaeran shortfall: consequences of using imputed databases”