Re: Group statistics for wavelets

TideMan <mulgor@gmail.com> wrote in message <e9e87c91-b2c0-49d4-8156-1908fb0ffe28@googlegroups.com>...
> On Thursday, January 30, 2014 7:14:09 AM UTC+13, M wrote:
> > I have 2 groups: A & B. Each subject in the group produces a signal as a function of time. I'd like to analyze these signals using wavelets and then compare the wavelets group A to group B, to see if, on average, they have different signals. Does anyone have any experience with doing this or can provide a reference for doing this? I appreciate any help that you can provide.
> >
> >
> >
> > Mike
>
> When you decompose using orthogonal wavelets you get a set of n details at timescales proportional to 2.^[1:n] and the approximation (i.e. remainder).
> Then to compare the signals, you can:
> * plot each detail from A with each detail from B;
> * calculate statistical moments of each detail;
> * cross-correlate each detail;
> * etc
> * etc

Thank you for your answer, but, I don't understand. If I create a wavelet of subject 1, I get a wavelet with a huge number (thousands) of coefficients, and from subjects 2 through 50, I again get wavelets with the same huge number of coefficients. Do I just take the mean of each of these coefficients, then do the same for the 50 wavelets from the Group B subjects? Can I take means of each of these coefficients? Are the means meaningful of anything? If I take the inverse of the mean coefficients, do I get the mean wave of the subjects? When I compare thousands of coefficients, by t-test, do I adjust the t-test for all these comparisons?
Thank you