help
81 Subscribers 
help > longitudinal analysis.
Showing 1-5 of 5 posts
Jul 23, 2020 10:07 PM | Xuehu Wei
longitudinal analysis.
Dears
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
Jul 24, 2020 08:07 AM | Andrew Zalesky
RE: longitudinal analysis.
Hi Xuehu,
You may want to check out some of the other posts on this forum that discuss the design matrix for longitudinal designs.
In brief, your design matrix would comprise:
1. A separate column for each subject. Each of these columns would include two ones corresponding to the two visits for that subject, while all other entries would be zero
2. A single column of ones and zeros indicating the time of visit (first=0, second=1)
3. An additional column for language score
4. The interaction between language score and time. You would element-wise multiply Columns 2 and 3 to get the interaction term. Note that you would need to remove column 3 if you want to study the interaction.
Another approach is to simply subtract structural connectivity at the first visit from structural connectivity in the second visit. You could then run the NBS on the DIFFERENCE in connectivity. This can be done with a simple design matrix comprising the language score (or it difference) and a column of ones.
Andrew
Originally posted by Xuehu Wei:
You may want to check out some of the other posts on this forum that discuss the design matrix for longitudinal designs.
In brief, your design matrix would comprise:
1. A separate column for each subject. Each of these columns would include two ones corresponding to the two visits for that subject, while all other entries would be zero
2. A single column of ones and zeros indicating the time of visit (first=0, second=1)
3. An additional column for language score
4. The interaction between language score and time. You would element-wise multiply Columns 2 and 3 to get the interaction term. Note that you would need to remove column 3 if you want to study the interaction.
Another approach is to simply subtract structural connectivity at the first visit from structural connectivity in the second visit. You could then run the NBS on the DIFFERENCE in connectivity. This can be done with a simple design matrix comprising the language score (or it difference) and a column of ones.
Andrew
Originally posted by Xuehu Wei:
Dears
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
Jul 24, 2020 09:07 AM | Xuehu Wei
RE: longitudinal analysis.
Dear Andrew,
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu
riginally posted by Andrew Zalesky:
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu
riginally posted by Andrew Zalesky:
Hi Xuehu,
You may want to check out some of the other posts on this forum that discuss the design matrix for longitudinal designs.
In brief, your design matrix would comprise:
1. A separate column for each subject. Each of these columns would include two ones corresponding to the two visits for that subject, while all other entries would be zero
2. A single column of ones and zeros indicating the time of visit (first=0, second=1)
3. An additional column for language score
4. The interaction between language score and time. You would element-wise multiply Columns 2 and 3 to get the interaction term. Note that you would need to remove column 3 if you want to study the interaction.
Another approach is to simply subtract structural connectivity at the first visit from structural connectivity in the second visit. You could then run the NBS on the DIFFERENCE in connectivity. This can be done with a simple design matrix comprising the language score (or it difference) and a column of ones.
Andrew
Originally posted by Xuehu Wei:
You may want to check out some of the other posts on this forum that discuss the design matrix for longitudinal designs.
In brief, your design matrix would comprise:
1. A separate column for each subject. Each of these columns would include two ones corresponding to the two visits for that subject, while all other entries would be zero
2. A single column of ones and zeros indicating the time of visit (first=0, second=1)
3. An additional column for language score
4. The interaction between language score and time. You would element-wise multiply Columns 2 and 3 to get the interaction term. Note that you would need to remove column 3 if you want to study the interaction.
Another approach is to simply subtract structural connectivity at the first visit from structural connectivity in the second visit. You could then run the NBS on the DIFFERENCE in connectivity. This can be done with a simple design matrix comprising the language score (or it difference) and a column of ones.
Andrew
Originally posted by Xuehu Wei:
Dears
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
I am new to NBS platform and wanted to use it for my longitudinal analysis.
I have 20 subjects, each subject was scanned twice( tp1: before training, tp2:after training).
a language score was recorded during these two visits.
Now I want to run a longitudinal analysis:
1. investigate the correlation between structural connectivity change and the language score change.
2. investigate the affection of the language score on the structural connectivity
Can you please advise how to write the design matrix?thank you very much.
Best
Xuehu
Jul 24, 2020 11:07 AM | Xuehu Wei
RE: longitudinal analysis.
Dear Andrew,
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu
Jul 24, 2020 11:07 PM | Andrew Zalesky
RE: longitudinal analysis.
Hi Xuehu,
this looks ok. It seems that the language score changes between the first and second visit for each subject. Therefore, there is not need to remove this column for language score. (I thought that the language score was the same for both time points. Disregard my previous advice about removing this column.)
[0 0 0 0 1 0] and [0 0 0 0 -1 0] is the main effect of language on connectivity
[0 0 0 0 0 1] and [0 0 0 0 0 -1] is the interaction between language and time on connectivity.
The exchange block is ok.
You may want to demean the language score.
Andrew
Originally posted by Xuehu Wei:
this looks ok. It seems that the language score changes between the first and second visit for each subject. Therefore, there is not need to remove this column for language score. (I thought that the language score was the same for both time points. Disregard my previous advice about removing this column.)
[0 0 0 0 1 0] and [0 0 0 0 -1 0] is the main effect of language on connectivity
[0 0 0 0 0 1] and [0 0 0 0 0 -1] is the interaction between language and time on connectivity.
The exchange block is ok.
You may want to demean the language score.
Andrew
Originally posted by Xuehu Wei:
Dear Andrew,
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu
Thank you very much, the follow is the simply design:
1 0 0 0 25 0
0 1 0 0 21 0
0 0 1 0 33 0
1 0 0 1 90 90
0 1 0 1 83 83
0 0 1 1 96 96
the first 3 columns for each subject , the 4th column for timepoint, 5th column for language score,
6th column for interaction between language score and time.
if I want to study the correlation between the connectivity change and score change,
the contrast : [0 0 0 0 0 1]
and if I want to study the correlation between the connectivity and the language score,
the contrast :[0 0 0 0 1 0]
all this with ttest th=3.
exchange block [ 1 2 3 1 2 3] .
Is this design accurate? because I really confuse on to remove score' column, if I want to study the interaction, remove score' column, and just keep last column how it calculate the change of language score.
Best
Xuehu