help > Design matrix
Showing 1-6 of 6 posts42 Subscribers
replyDisplay: Results per page:
May 10, 2017  04:05 AM
Design matrix
Dear all,

I would like to run a comparison between two groups adjusted for covariates. I have 4 covariates, the first two are continuous and the second 2 are binary. The design matrix that I have is:

1 0 -13.5490196 7.4118 -0.27451 -0.294118
1 0 -10.5490196 -5.5882 0.72549 0.705882
1 0 16.4509804 15.4118 -0.27451 -0.294118
1 0 -1.5490196 15.4118 -0.27451 0.705882
0 1 -9.5490196 8.4118 -0.27451 -0.294118
0 1 -8.5490196 -4.5882 0.72549 -0.294118
0 1 -8.5490196 -4.5882 -0.27451 -0.294118
0 1 14.4509804 13.4118 -0.27451 -0.294118
0 1 23.4509804 -1.5882 -0.27451 -0.294118

And the contrast I use:
[1 -1 0 0 0 0]

I don't know if this is correct, any advice will be really helpful. Thanks in advance.

Regards,

Manuel
reply reply with quote
May 10, 2017  07:05 PM
RE: Design matrix
Hi Manuel,

the design matrix and contrast looks correct. Given that you only have 9 subjects/samples, you might want to consider omitting some of the covariates, but there is nothing wrong with the current design.

Andrew

Originally posted by Manuel Blesa:
Dear all,

I would like to run a comparison between two groups adjusted for covariates. I have 4 covariates, the first two are continuous and the second 2 are binary. The design matrix that I have is:

1 0 -13.5490196 7.4118 -0.27451 -0.294118
1 0 -10.5490196 -5.5882 0.72549 0.705882
1 0 16.4509804 15.4118 -0.27451 -0.294118
1 0 -1.5490196 15.4118 -0.27451 0.705882
0 1 -9.5490196 8.4118 -0.27451 -0.294118
0 1 -8.5490196 -4.5882 0.72549 -0.294118
0 1 -8.5490196 -4.5882 -0.27451 -0.294118
0 1 14.4509804 13.4118 -0.27451 -0.294118
0 1 23.4509804 -1.5882 -0.27451 -0.294118

And the contrast I use:
[1 -1 0 0 0 0]

I don't know if this is correct, any advice will be really helpful. Thanks in advance.

Regards,

Manuel
reply reply with quote
May 12, 2017  06:05 AM
RE: Design matrix
Hi,

Thanks, that's really helpful, I will consider removing some of them.

The experiment has more subjects (51) but here I only wrote a small piece of the matrix.

Best Regards,

Manuel
reply reply with quote
May 12, 2017  07:05 PM
RE: Design matrix
Given that you have 50 subjects, omitting the covariates is probably unwarranted. I wrongly assumed you only had 9 subjects.
Andrew

Originally posted by Manuel Blesa:
Hi,

Thanks, that's really helpful, I will consider removing some of them.

The experiment has more subjects (51) but here I only wrote a small piece of the matrix.

Best Regards,

Manuel
reply reply with quote
Oct 4, 2017  01:10 AM
RE: Design matrix
Hi,

I have a final doubt about the design matrix. Mi interest is to compare two groups and see if there show differences, the connectivity matrix is the same as before:

1 0 -13.5490196 7.4118 -0.27451 -0.294118
1 0 -10.5490196 -5.5882 0.72549 0.705882
1 0 16.4509804 15.4118 -0.27451 -0.294118
1 0 -1.5490196 15.4118 -0.27451 0.705882
0 1 -9.5490196 8.4118 -0.27451 -0.294118
0 1 -8.5490196 -4.5882 0.72549 -0.294118
0 1 -8.5490196 -4.5882 -0.27451 -0.294118
0 1 14.4509804 13.4118 -0.27451 -0.294118
0 1 23.4509804 -1.5882 -0.27451 -0.294118

(it's only a sample, the full matrix has 47 subjects)

This matrix has been used in TBSS as well, the question is which matrix should I use? this is the demeaned matrix, or maybe is better to use the raw matrix?

For example this:

1 0 27.57 39.43 1 1
1 0 27.43 40.86 0 0
1 0 27.86 39.00 1 1
1 0 31.71 42.00 0 0
1 0 29.14 42.00 1 0
0 1 28.00 41.00 0 0
0 1 28.14 39.14 0 0
0 1 31.43 41.71 0 0
0 1 32.71 39.57 0 0
0 1 28.71 38.43 0 0
0 1 31.00 39.00 0 0

And one final question: at the time of choosing the statistical method, I use t-test with threshold of 3, is the appropiate test to compare group differences? and how I can interpret this threshold? Thanks in advance.

Regards,

Manuel
reply reply with quote
Oct 4, 2017  08:10 AM
RE: Design matrix
Hi Manuel,

the first design matrix has fewer subjects than the second one, and thus they are very likely to yield different results.

In general, demeaning the covariates will not make a difference if you are modelling the mean as part of the main effect.

The t-test is appropriate to compare groups. Further information about the threshold can be found in the manual.

Andrew


Originally posted by Manuel Blesa:
Hi,

I have a final doubt about the design matrix. Mi interest is to compare two groups and see if there show differences, the connectivity matrix is the same as before:

1 0 -13.5490196 7.4118 -0.27451 -0.294118
1 0 -10.5490196 -5.5882 0.72549 0.705882
1 0 16.4509804 15.4118 -0.27451 -0.294118
1 0 -1.5490196 15.4118 -0.27451 0.705882
0 1 -9.5490196 8.4118 -0.27451 -0.294118
0 1 -8.5490196 -4.5882 0.72549 -0.294118
0 1 -8.5490196 -4.5882 -0.27451 -0.294118
0 1 14.4509804 13.4118 -0.27451 -0.294118
0 1 23.4509804 -1.5882 -0.27451 -0.294118

(it's only a sample, the full matrix has 47 subjects)

This matrix has been used in TBSS as well, the question is which matrix should I use? this is the demeaned matrix, or maybe is better to use the raw matrix?

For example this:

1 0 27.57 39.43 1 1
1 0 27.43 40.86 0 0
1 0 27.86 39.00 1 1
1 0 31.71 42.00 0 0
1 0 29.14 42.00 1 0
0 1 28.00 41.00 0 0
0 1 28.14 39.14 0 0
0 1 31.43 41.71 0 0
0 1 32.71 39.57 0 0
0 1 28.71 38.43 0 0
0 1 31.00 39.00 0 0

And one final question: at the time of choosing the statistical method, I use t-test with threshold of 3, is the appropiate test to compare group differences? and how I can interpret this threshold? Thanks in advance.

Regards,

Manuel
reply reply with quote
reply