help
help > RE: How to use "age" as covariates?
May 15, 2014 01:05 PM | Yifei Zhang
RE: How to use "age" as covariates?
Hi Alfonso,
Thank you very much for the detailed information!
If I want to do the group comparison between patient and control groups for the same regression (Functional connectivity (small-world measurements)~ All(intercept)+A+gender+B), but to check the association between A and the small-world measurement(A as predictor). I am not sure if I did it the right way as follows:
1) Define the 2nd-level covariates: Patient(1 for patients, 0 for controls), Control(1 for controls, 0 for patients), A_patient (A values for Patients, 0 for controls), A_control (A values for controls, 0 for patients), Gender and B.
2) a) select all of the covariates above in the 2nd-level analysis window and use [0 0 1 -1 0 0] as contrast.
b) select Patient, Control, A(a new covariate does't separate to two groups), gender and B, and use the contrast of [1 -1 1 0 0].
I am not sure which one of the two definitions is the right one.
3) check the result in "Graph theory results explorer", try to use different network thresholds of cost like 0.3-0.5 (is there any suggestion on the range of this threshold for the analysis between HC and AD groups?). Then the group comparison could be explained by checking the beta value of the significant ROIs. If the beta is positive, that means the association between A and the certain small-world measurement(e.g. Global efficiency) in the patient group is larger than in the control group?
Am I right for the steps above?Additionally, if I check the results in ROI-to-ROI results explorer, I am not quite understand what's the difference between the two thresholds of p-FDR(seed-level correction) and the p-FDR(analysis-level correction) in the ROI-to-ROI connections(by intensity).
According to the No. 16 thread of the FAQ in this forum. The p-FDR(analysis-level correction) is the most strictly one for individual ROI-to-ROI connections. If we want to obtain the inference about which ROIs show significant effects we could use cases (b) and (c) methods, which are the F-test and NBS method, by defining the extent(seed- or network- level) thresholds. I don't understand how to make inferences of the p-FDR(seed-level correction) threshold in the first line(ROI-to-ROI connections(by intensity)).
I have carefully read the FAQ and other information in this forum, and have discussed the "2) question" above with my colleague. We could not come to an agreement on that. I really what to hear from your opinions.
I appreciate a lot for your help!
Best regards,
Yifei
Thank you very much for the detailed information!
If I want to do the group comparison between patient and control groups for the same regression (Functional connectivity (small-world measurements)~ All(intercept)+A+gender+B), but to check the association between A and the small-world measurement(A as predictor). I am not sure if I did it the right way as follows:
1) Define the 2nd-level covariates: Patient(1 for patients, 0 for controls), Control(1 for controls, 0 for patients), A_patient (A values for Patients, 0 for controls), A_control (A values for controls, 0 for patients), Gender and B.
2) a) select all of the covariates above in the 2nd-level analysis window and use [0 0 1 -1 0 0] as contrast.
b) select Patient, Control, A(a new covariate does't separate to two groups), gender and B, and use the contrast of [1 -1 1 0 0].
I am not sure which one of the two definitions is the right one.
3) check the result in "Graph theory results explorer", try to use different network thresholds of cost like 0.3-0.5 (is there any suggestion on the range of this threshold for the analysis between HC and AD groups?). Then the group comparison could be explained by checking the beta value of the significant ROIs. If the beta is positive, that means the association between A and the certain small-world measurement(e.g. Global efficiency) in the patient group is larger than in the control group?
Am I right for the steps above?Additionally, if I check the results in ROI-to-ROI results explorer, I am not quite understand what's the difference between the two thresholds of p-FDR(seed-level correction) and the p-FDR(analysis-level correction) in the ROI-to-ROI connections(by intensity).
According to the No. 16 thread of the FAQ in this forum. The p-FDR(analysis-level correction) is the most strictly one for individual ROI-to-ROI connections. If we want to obtain the inference about which ROIs show significant effects we could use cases (b) and (c) methods, which are the F-test and NBS method, by defining the extent(seed- or network- level) thresholds. I don't understand how to make inferences of the p-FDR(seed-level correction) threshold in the first line(ROI-to-ROI connections(by intensity)).
I have carefully read the FAQ and other information in this forum, and have discussed the "2) question" above with my colleague. We could not come to an agreement on that. I really what to hear from your opinions.
I appreciate a lot for your help!
Best regards,
Yifei