Missing inaction: the dangers of ignoring missing data

摘要

The most common approach to dealing with missingdata is to delete cases containing missing observations.However, this approach reduces statistical power andincreases estimation bias. A recent study shows howestimates of heritability and selection can be biasedwhen the 'invisible fraction' (missing data due tomortality) is ignored, thus demonstrating the dangersof neglecting missing data in ecology and evolution. Wehighlight recent advances in the procedures of handlingmissing data and their relevance and applicability.