Of one thing I am certain: count data should be integers (or half integers if partial individuals are counted as one half). The chironomid data from Lake Żabińskie are remarkable in many respects; surely they meet this most basic expectation.
#load the data library(readxl) fname <- "zabinskie2015cit.xls" excel_sheets(fname) fos_counts <- read_excel(fname, sheet = "Chironomids Zabinskie counts") chron <- fos_counts[, 1] fos_counts <- fos_counts[, -c(1, ncol(fos_counts))]
Let’s look at the top left corner of the data matrix
|Chironomini||Chironomus anthracinus||Chironomus plumosus|
In the fossil data set, 2.5% have non-integer (or half integer) values. These non-integer counts have a probabilty of occuring of exactly zero. Cromwell’s rule is over.
We can be absolutely certain that the archived data are not the original counts. Most likely, they are back-transformed from percent data that had been rounded. Why would anybody do that? Perhaps the authors lost the original counts. That would be unfortunate.