Are you using the same aggregation windows and same number of repeats for the gait score and the individual features?

I discussed this with Daria and Stan before. The baselines for clinical and digital measures are 6-8 weeks apart. While we know that there is some learning effect even for the gait features I would like us to consider the classical baseline. Use the first 2 weeks or maybe ignore the first week and consider the 2nd and 3rd week as your baseline and re-do these analyses. I suspect you will see a bit more movement on the FL features if you do this.

This group actually include two (or three) types of patients: - Patients who worsen <20% - Patients who are stable (no change at all - probably not that many) - Patients who improve

There are ways to separate those by using e.g. latent class trajectories modelling

Hi Vittorio, can you please confirm what are you doing here. You look at patients with >20% increase in 25FWT, and you define those as progressors. You then look at change in 25FWT (mean?) i that group and in the remaining patients (non-progressors)?