Hi, Dan-
OK, thanks for posting that. So when you suggest that the "initial signal is unstable", it's not like the huge pre-saturation values (which presumably there would have been warnings about)? Instead, it's a more subtle, yet still correlation-driving pattern, it seems. Perhaps that makes sense that it *wasn't* a pre-steady state spike, since I think you used subbrick selectors to make each time series from a single run, so it couldn't have been two matching pre-steady state spikes or something (and that might have shown up globally, too).
Anyways, I was just curious. I'm glad you sorted out the issue.
Thanks,
pt