Biotech Values

[dewophile]

To elaborate on Dew's explanation

emis took continuous data (hgb A1C), and instead of running the appropriate statistics, arbitrarily picked a cutoff at their discretion (that is not to say 1.1% diff is not clinically meaningful, but there is no formal 1.1% cutoff for clinical, regulatory or other purposes that is ever used), simply to show statistical significance on some parameter. in this case they essentially reduced mroe powerful parametric statistics to a binary non-parametric test.
The inference is that NO statistical significance in any group would be seen with appropriate parametric statistics using the raw data, and picking a cutoff at one's discretion from a body of data jsut to show statistical significance is not valid, particularly from a small subset in which random events can trigger statistical differences if enough comparisons (in this case varying "cutoffs") are analyzed
at least that is what I think Dew is saying