As you might remember in this post we saw an interesting characteristic in the visual fields. I got in touch with Koen Vermeer from the Rotterdam Ophthalmic Institute and he was kind enough to inform me about a small problem with the dataset that is currently solved.
February 18, 2015 - Problem found (and fixed) in longitudinal glaucomatous visual field data set
An observation by Angel de la Llama on his EyeData blog triggered us to revisit the longitudinal visual field data set. As it turned out, there was a mix-up in some of the IDs of the visual fields. This does not affect which visual fields are included - only their ID was affected. This problem concerned less than 4% of the visual fields. We advise you to download the latest version of the data set and rerun your analyses.
So I downloaded the latest version (they have a new website for the data by the way) and rerun the code as advised and here is the result:
That is more what I had expected. Now there is only one point per eye gaining sensitivity on average and actually very little (0.2 dB for OD and 0.8 dB for OS).
Mystery solved !