Initial SLAP2 QC analysis

[ido4848]

Here's an initial QC analysis I did on the slap 2 data, mainly trying to compute the SNR:

The relevant notebook is here: https://github.com/AllenNeuralDynamics/openscope-community-predictive-processing/blob/aizenbud-slap2-snr/docs/notebooks/pilot_analysis_signal_qc.ipynb

I would be happy to hear your comments, as this is my first time working with such data.

[maierav]

Very interesting. Out of these measures, z-score might be the metric most people already are most familiar with.

I am not quite sure I fully grasped "Reliability" - it sounds like a within-trial comparison? Maybe worth adding cross-trial STD or SEM?

Coming from spiking data, those are the two SNR-related measures I would be most interested in:
(i) Response magnitude (z-score is great; percent change also has its advantages)
(ii) Response variance to repeated presentations (STD or SEM across trials/stimulus presentations)

The reason is that even a response with a low z-score is interesting, if consistent across stimulus repetitions. On the flip side, a high z-score is less interesting if the z-score varies a lot between stimulus presentations.

[ido4848]

Thanks for the feedback.

See the following update:

response_t_stat was defined as response_z_mean / response_z_sem to capture the intuition of a low z-score while remaining consistent.

[maierav]

Awesome. I was about to suggest t-scores.

Now we could even transform them into p-values by using the sample size on a t-distribution. Then, we could shade the histogram by statistically significant versus not significant values (p<0.05). Might help make it more intuitive to readers (though, personally, I am not a fan of that arbitrary threshold).

This is one session, correct? What might be a good way to summarize across the full dataset (more sessions are incoming from what I understand)?