search

zqwei / Zebrafish_spinal_cord_development

8 stars 3 forks source link

Analysis the data after baseline recomputation #5

Closed zqwei closed 7 years ago

zqwei commented 7 years ago

Remove the fake hyperpolarization in calcium data while using alg 1 in baseline computation

zqwei commented 7 years ago

@yinan-wan Done for the all datasets --- alg 2

Plots: https://www.dropbox.com/sh/x4jsaij6nyd23h0/AACKj2D_GK0nNt5Jw2cWNfbCa?dl=0 TempDat: https://www.dropbox.com/sh/n8rc353wqox8ng7/AACuljHyJEQYSwH3tvoqZF7Ga?dl=0

zqwei commented 7 years ago

Other tests are listed in wikis

https://github.com/zqwei/Zebrafish_spinal_cord_development/wiki

zqwei commented 7 years ago

@yinan-wan the results based on the alg 3 is done for dataset 150410.

https://www.dropbox.com/sh/jlfeii62ddf6ukq/AAC_WBYb6OjFeMESyIbX7GGua?dl=0

Although one can still see the slow fluctuation, this effect is hugely reduced. I think some of these 'fake factors' will influence your decision for pioneer neurons, since one neuron could be active due to the slow fluctuation, but contain no signal. Probably, it is better for you to go through the data and figure first before we continue with the other datasets.

zqwei commented 7 years ago

@yinan-wan also I went back to see the results from alg 1, there are smaller number of neurons with slow fluctuation, but they are barely factored. However, both using alg 2/3 the neuron with slow fluctuation are intended to be factored. I guess the slow fluctuation on some degree is global across many neurons which forms a fake factor that contains no signal at all. We should avoid these.

zqwei commented 7 years ago

@yinan-wan Done for alg 2, w = 61, p = 20% with joint of multiple activation mats.

  1. alg 1 baseline
  2. alg 2 baseline, p = 20%, w from 21, 41, 61

It seems slightly different in a few frame comparing to the old alg 1 results. Please have a look if that is acceptable or we need a further discussion.

yinan-wan commented 7 years ago

@zqwei active neurons are indeed different because extra filtering is done for activation test. This leads to very clean factors and sometimes may be suppressing active neurons with very weak signals. It looks different, but I won't say it's wrong. Key point is weather early local factors become unstable. If it is not the case, I'll say it's fine.