Rank MGC

[cshen6]

We should re-visit this too, post-submission.

Now we understand much better than 1 year ago, we can do a better rank MGC now. I tried briefly, some interesting findings:

1. With a few tweaks, rank MGC is almost equivalent throughout dcorr / mcorr / MGC, theoretically and numerically. So it can be a single framework.
2. In existing simulations, rank MGC is equivalent to MGC_{m} for almost all simulations, including its h-d advantages.
3. For real data, rank MGC can now detect significant p-value for Brain CxP.

So we should work on this post-submission too, and include it in the grant application.

[jovo]

agreed. also, MGC detects significant p-value for Brain CxP, but you have to use my "p = min t s.t. a cc w/|cc|>3 exists" strategy

On Sun, Jul 3, 2016 at 11:36 AM, cshen6 notifications@github.com wrote:

We should re-visit this too, post-submission.

Now we understand much better than 1 year ago, we can do a better rank MGC now. I tried briefly, some interesting findings:

1. With a few tweaks, rank MGC is almost equivalent throughout dcorr / mcorr / MGC, theoretically and numerically. So it can be a single framework.
2. In existing simulations, rank MGC is equivalent to MGC_{m} for almost all simulations, including its h-d advantages.
3. For real data, rank MGC can now detect significant p-value for Brain CxP.

So we should work on this post-submission too, and include it in the grant application.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/jovo/MGC/issues/73, or mute the thread https://github.com/notifications/unsubscribe/AACjcqkiUNhPHgQV8Y1TTeo7TFVQpx9dks5qR9b1gaJpZM4JD5DA .

the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York

[cshen6]

I will move this issue to the other repo