Hi Steve,

I can't check it right now but this looks right to me. We'd normally think on the sums of squares instead of variances, but since the N is the same across all the model, this shouldn't matter.

As a quick check, you can verify that var(Y) is the same as the sum of the vars for each of the EVs computed separately using the equation, plus the variance of the residuals.

All the best,

Anderson

On 19 December 2013 15:58, Steve Mayhew <[log in to unmask]> wrote:

Hi guys,

I want to calculate the % variance explained by each of my GLM regressors for a given voxel coordinate (x,y,z).

From looking through old posts i found this equation, but just wanted to check ive got it right and im using the right files.

a^2 var(EV)
----------------
var(Y)

Here data = Y, a = PE

So I square the PE value for that voxel, multiply it by the variance of the regressor timeseries taken from the design.mat and then divide the result by the variance of the filtered_func_data timeseries in that voxel?

Many thanks for any help
Steve