Task fMRI Simulations: Calculate Signal to Noise Ratio


Hi All!

I’m working on making simulations similar to Jeanette Mumford et al.'s 2012 paper. I’m curious how the signal-to-noise ratio was calculated, and how large I should make my task activation relative to the noise to reflect “real” data. I found this quote in the paper, but I still feel uncertain how signal-to-noise was actually calculated.

Twelve sets of simulations were carried out, one for each combination of ISI interval ( umin, umax ) and signal-to-noise level (determined by σ2)

@Jeanette_Mumford, would you have time to answer these questions?

You can pull up and play with my notebook here



Hi! I just used what values I cited as sigma in the equation just above that. I don’t think I actually calculated the SNR, but was kept a fixed beta magnitude (also a little above that).

Did you check out this paper?

Good one and they improved my simulations a bit. If memory serves I should have scaled my regressor height after convolution and I didn’t.
There’s a related 2014 paper with similar situations. Code is here

Hope this helps! Feel free to email me directly if you have more questions. Use my gmail.


1 Like