I do not have a strong background in statistics, so am unable to follow all the reanalysis. So I decided to run my own much simplified analysis on the source data spreadsheet. What is confusing me is I am getting lower participant recovery numbers for original endpoint than the reanalysis did. e.g. 6 for CBT rather than 11, 5 for GET rather than 7, 1 for APT rather than 3, and 2 for SMC rather than 5. I do not think (but I am by no means sure) this is simply due to my non-statistical approach, but would like to know one way or the other.
I have only counted participants who fulfilled all four criteria. Because the original protocol endpoint ME/CFS criterion is not available, I used the simple Oxford criterion that is available, meaning my numbers would still give recovery rates on the optimistic side.
If anyone can help clarify why I am seeing different numbers I would appreciate it. Happy to make my spreadsheet available if need be.
I have some basic understanding of statistics, and would be grateful to better understand the stats techniques used in the reanalysis. Any pointers would be much appreciated.