search

Eggeling-Lab-Microscope-Software / TRAIT2D

TRAIT2D is a cross-platform Python software package with compilable graphical user interfaces (GUIs) to support Single Particle Tracking experiments.
GNU General Public License v3.0
10 stars 10 forks source link

Time step size dependency in models #28

Closed john-wigg closed 2 years ago

john-wigg commented 4 years ago

I was wondering if it was possible to formulate the the models for brownian, confined and hopping diffusion independently of the fixed time step size dt.

Right now, the models contain i which can be written as i = t/dt where t is the time.

Since smart averaging now works regardless of any time step size (see PR #27) this is the last piece of code in the analysis library that still depends on the tracks having a fixed stepsize.

FReina commented 4 years ago

This is in principle possible. However the issue is, how would the averaging, to obtain amean squared displacement, work in this scenario? Because we would have to make a choice of how points get grouped. An example to clarify: let's suppose that our points are all at a distance of 1s, apart from one localization that happens at 1.2. So this point would be an odd one out and it would be better to merge it to another set of displacements to average, with the caveat that now there is a time uncertainty of 0.2 seconds (well not quite, but you get what I mean).

This will be interesting for us in the (hopefully near) future, but it is a bit beyond the scope of the initial release.

john-wigg commented 3 years ago

For reference: The Minflux branch, which supports variable time steps, handles this by using the smallest time step throughout the track.

FReina commented 2 years ago

Indeed, this is handled in the Minflux branch on my fork. The aim would be to generalize the calculations for non-constant time intervals.

I will add this to the Roadmap for v.2, reference this issue and close it for now.