Skip to content Skip to navigation

Dependencies and contingencies

Saturday, February 1, 2020 - 10:25am
"Dependencies and contingencies"

Count data are often recorded and analysed as contingency tables whose within-row or within-column categories are considered independent. In recent work on the human microbiome, we have come across cases where all categories present dependencies invalidating the traditional “degrees of freedom” and relevant decompositions. For instance if the rows are different species of bacteria whose phylogenetic relationships are known, supposing that this classification is unknown discards important information. The framework developed by Persi and his co-authors where tables come with given row sums and column sums does not apply here and I will present some more realistic approaches that attempt to overcome this challenge.