Print

Print


Hi Anderson,

Thank you so much for your explanation. I think I'm just struggling with using Text2Vest to generate the number of waves/points that I need. Specifically, I am trying to correlate Fa values to a single behavioral measure so need my design matrix to be 100X1, and my contrast matrix to simply to represent the single EV. 

However, I cannot find a format in which Text2Vest will probably create the contrast file for me. Just a 1 for a single contrast will give me a division by 0 error and a 1 0 gives me a 1X2 matrix which is not what I need. 

Thank you so much,
Levi 

On Sat, Oct 17, 2015 at 4:38 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi,

Maybe that example was too tiny to be useful. The file contains a matrix. Each row of this matrix is a t-contrast.

The /NumWaves is the number of columns in this matrix (which corresponds to the number of EVs in the design matrix), and /NumPoints is the number of rows (which corresponds to the number of t-contrasts).

If you want a single contrast as [1 -1], that is, comparing the regression coefficients for EV1 with EV2, the contrast file would be:

/NumWaves 2
/NumPoints 1

/Matrix
1 -1

All the best,

Anderson


On 16 October 2015 at 17:28, LEVI SOLOMYAK <[log in to unmask]> wrote:
Hi Anderson,

Thank you so much for your suggestion. When I try and create the contrast file from that matrix (1, -1), it interprets it as NumWaves 2 and NumPoints 1. Is there some simple way so that it defines the contrast the other way?

Thank you so much,
CN 

On Fri, Sep 25, 2015 at 2:15 AM, Anderson M. Winkler <[log in to unmask]> wrote:
Hi Corey,

Can't say without seeing what you are doing, but in any case, with a single EV, you can create the design.mat by hand. It's a text file that looks like this:

/NumWaves 1
/NumPoints 100

/Matrix
xx
xx
xx
xx
xx
...

Where the "xx" are the numerical values you have for the EV. The design.con follows a similar structure:

/NumWaves 1
/NumPoints 2

/Matrix
1
-1

When running randomise, include the option -D.

All the best,

Anderson



On 24 September 2015 at 16:38, Corey N <[log in to unmask]> wrote:
Dear FSl Experts,

I am hoping to run a one sample t-test using GLM to find correlations between FA values and a behavioral measure. Since all of my subjects are in a single group, my design matrix is simply the behavioral value. Is that correct?

Because I am working with 100 subjects, I used Text2Vest to convert my contrast matrix. I tried to do the same for the design matrix (1) since theres only 1 ev but get a expr: division by zero.

Am I setting up my GLM correctly? And if so, what is my mistake with Text2Vest?

Thank you so much
CN