A Cautionary Note on Phylogenetic Signal Estimation from Imputed Databases

摘要: Given the prevalence of missing data on species’ traits – Raunkiaeran shortfall-, several methods have been proposed to fill sparse databases. However, analyses based these imputed databases can introduce biases. Here, we evaluated potential estimation biases caused by use In evaluation, considered descriptive statistics, regression coefficient, and phylogenetic signal for different imputing scenarios. We found that percentage data, mechanisms imputation were important in determining errors. Imputation errors are not linearly related estimate Adding information provides better estimates but this should be combined with other variables such as correlated variable. Using an empirical dataset, even strongly each other, brain body size primates, produce when estimating from datasets. advise researchers share both their raw well consider pattern evaluate perform goals. addition, performance mainly statistical instead only error.