search

stephlj / Traces

Analysis software for single-molecule fluorescence resonance energy transfer (smFRET) data.
MIT License
9 stars 1 forks source link

compare custom trans-fitting code to Matlab's fitgeotrans #4

Closed mattjj closed 10 years ago

mattjj commented 10 years ago

Quoth the stephie:

Perhaps you can run the channel mapping half with the "Good Beads" data set and look at the A,b that I calculate using the algorithm you showed me, compared to Amatlab and bmatlab obtained with the built-in fitgeotrans? As noted in the README, fitgeotrans does better if you want to overlay the images (which is good cuz that's what it's made for), but my/our version does better when you want to map a particular pixel from one channel to another.

I plotted various things and can see what's different about the two transformations, but not why one works better for image overlay and the other for point-to-point mapping.

mattjj commented 10 years ago

Is this still an open issue, or should it be closed?

stephlj commented 10 years ago

going to close it but open a related issue about overlays

mattjj commented 10 years ago

Btw the most fun thing is that github will auto-close issues if you mention them appropriately in a git commit that you push. Like if you push a commit with a message that says fixes #4 it will close the corresponding issue.

Not relevant here but I thought you should know that it's the most fun thing.