Print

Print
Dear Vladimir,
 
If X is my grip strength vector and Y is my stimuli onset.

I tried spm_orth as : X=spm_orth(X)

I think I am missing something as I need it to be with respect to Y. When I hit enter, it gives me the same result. 

I downloaded spm12 and it has the orthogonality option and I think it did it right as in the viewing it says orthogonal under the columns. However, I need to confirm few things: what is the best polynomial order I should use. I used the fist and it results in three additional regressors. Do I need to mean corrected? 

Then, in spm, how you can see/view the effects of the parameter in the fitted response? And is it possible to see in spm the effects before and after the parameter modulation

Thank you very much

AS

On 29 Apr 2013, at 13:28, Vladimir Bogdanov <[log in to unmask]> wrote:

Dear AS,

This could be helpful:

https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;1de7f347.1103

You might consider SPM function smp_orth.

Sincerely yours,
Vladimir

Volodymyr B. Bogdanov, PhD

--- On Sat, 4/27/13, AS T.N <[log in to unmask]> wrote:

From: AS T.N <[log in to unmask]>
Subject: [SPM] Investigating the effects of Parameters
To: [log in to unmask]
Date: Saturday, April 27, 2013, 5:43 PM

Dear friends ,


I will start explaining briefly the experiment that I am thinking of, so it is an example, then asking my questions..

So if you are interested in, say, the hand area in the brain and you do a block design of 5 tasks and 5 rests. Each lasts for, say, 20s. At the same time you have a squeeze pressure or ball. The squeeze vales, or the grip strengths, recorded as a vector for each time the subject squeezes. Then, you design the interval time between each stimuli of squeeze as, say 1s. So you will have 20 squeezes per each task block. The squeeze values are recorded as percentages so for example 99 95 90 99....etc

Questions:

1) What is the best way to enter the condition in this case if I am interested in knowing the effects of the grip , or how much does the subject squeeze, in the neuronal activity..etc?  What I did is that I defined a vector and entered starting time of each stimuli as a delta function, just for the active task. So I would have a vector  Y   that contains 100 values, 20*5. Then I entered the duration as 0. So I model it as if it is an event design.


2) What is the best way to add the squeeze values in the model to see their effects on the result?. I have read and tried to add them under the parametric modulation option. However, I have read in different articles that they mean corrected and orthogonalised  such vector. I did the mean correction using this equation : if x  is our  grip strength vector : x_corrected= x-mean(x).  However, I am not sure about the orthogonality and how to do it correctly. because in SPM manual it says that X'*Y= 0 and the cosine =0. I used this equation to do the orthogonality :
X_orth = X - (X'*Y)/(Y'*Y)*Y;

Y is our condition time or active time. I am not sure whether it is correct to do it or not. However, in SPM when you see orthogonality it says it is not orthogonal nor collinear and it has a white box under its column. 

3) What is the best or accurate way to defined the contrast in this case ? What I did was that I actually tried more than one option. So at the end I had two columns ( regressors) , one for the active time and the other for the parametric modulation. I defined the contrast as [ 1 ] only so SPM will give the others 0. Or is it best to defined both of them [ 1 1 ]?

4) If I defined the parametric modulation or the grip strength in the contrast only, [ 0 1], in the result I get different active brain regions that do not correlate or are not in the same location of the hand area compared with what I get from the condition regressor. Is this actually right? or is what I am doing right of looking for such result. Or in other words, does including the parameter vector in my model as a regressor enough to see its effects even without defining them in the contrast, i.e, [ 1  0 ]? 

Thanks in advance

AS