help
help > RE: covariates (of interest and control) set-up
Nov 6, 2018 12:11 PM | therese1
RE: covariates (of interest and control) set-up
Originally posted by therese1:
Hello Alfonso,
I'd love to get your opinion on the questions
below. I'd greatly appreciate any help you can offer.
Many thanks,
Therese
Hello,
Please can you advise whether I have set up covariates correctly for a GLM 2nd level analysis.
Covariates of interest:
AllSubjects [1,1,1,1,1....1,1,1]
Patients [1,1,1,1.......0,0,0]
Controls [0,0,0,0.......1,1,1]
Test_1 [-6.75, 0.18, -4.25.....]
Test_2 [-0.63, 0.92, -1.46,.....]
Control covariates:
Age (demeaned) [3.16, 10.16, -9.56.....]
Education (demeaned) [-0.33, 4.66, -2.33,....]
Gender [-0.5, -0.5, 0.5, 0.5,....]
All covariates also have a sub-group version e.g. Test_1_Patients, Gender_Controls.
Question 1:
The test scores, for each test, are normalised based on a population sample matched for education, gender & age (i.e., the score for person 1 and person 2 may have been normalised based on different populations. Formula score - population mean / population SD). In this case zero is meaningful. Do these need to be further normalised at a whole group level?
Question 2:
Are the control convariates set up correctly? I have read conflicting items on how to set up gender.
Question 3:
I have selected semi-partial correlations for ROI-to-ROI investigation.
One analysis is set up as:
Test_1_Controls > Test_1_Patients; one-sided positive
[0,0, -1, 1, 0, 0, 0] referring to [Patients, Controls, Test_1_Patients, Test_1_Controls, Gender, Age, Education]
Is it correct to say that this set up is testing whether in relation to Test_1 controls show higher ROI-to-ROI functional connectivity compared to patients while controlling for gender, age and education?
Thank you in advance for any assistance.
Therese
Please can you advise whether I have set up covariates correctly for a GLM 2nd level analysis.
Covariates of interest:
AllSubjects [1,1,1,1,1....1,1,1]
Patients [1,1,1,1.......0,0,0]
Controls [0,0,0,0.......1,1,1]
Test_1 [-6.75, 0.18, -4.25.....]
Test_2 [-0.63, 0.92, -1.46,.....]
Control covariates:
Age (demeaned) [3.16, 10.16, -9.56.....]
Education (demeaned) [-0.33, 4.66, -2.33,....]
Gender [-0.5, -0.5, 0.5, 0.5,....]
All covariates also have a sub-group version e.g. Test_1_Patients, Gender_Controls.
Question 1:
The test scores, for each test, are normalised based on a population sample matched for education, gender & age (i.e., the score for person 1 and person 2 may have been normalised based on different populations. Formula score - population mean / population SD). In this case zero is meaningful. Do these need to be further normalised at a whole group level?
Question 2:
Are the control convariates set up correctly? I have read conflicting items on how to set up gender.
Question 3:
I have selected semi-partial correlations for ROI-to-ROI investigation.
One analysis is set up as:
Test_1_Controls > Test_1_Patients; one-sided positive
[0,0, -1, 1, 0, 0, 0] referring to [Patients, Controls, Test_1_Patients, Test_1_Controls, Gender, Age, Education]
Is it correct to say that this set up is testing whether in relation to Test_1 controls show higher ROI-to-ROI functional connectivity compared to patients while controlling for gender, age and education?
Thank you in advance for any assistance.
Therese
Threaded View
Title | Author | Date |
---|---|---|
therese1 | Oct 30, 2018 | |
therese1 | Nov 6, 2018 | |
therese1 | Nov 9, 2018 | |
Jeff Browndyke | Nov 9, 2018 | |