help
help > RE: Why does the same condition in two analyses produce different results?
May 27, 2020 10:05 PM | Alfonso Nieto-Castanon - Boston University
RE: Why does the same condition in two analyses produce different results?
Dear YJ
That's an interesting question. The only possible difference I can think of arises from different denoising procedures. Is it correct to assume that in your original analyses (with all the temporal conditions) you included all of the 51 'effect of rest x Time#" effects in the list of confounding effects during denoising (this is the proper/default thing to do for sliding-window temporal analyses), while in the latter analyses you did not (only a single 'effect of rest' effect was included)? If that is the case, then, depending on the window length and overlap, the former analyses would remove a lot of the low frequency BOLD signal variability compared to the latter analyses as part of denoising, so your resulting original "rest" connectivity is likely focusing on the contribution of higher-frequency effects while the latter "rest" connectivity analysis is focusing on a broader frequency range, which may explain the general similarities/differences between the two analyses.
Let me know if that makes sense and/or your thoughts
Alfonso
Originally posted by ys1j13:
That's an interesting question. The only possible difference I can think of arises from different denoising procedures. Is it correct to assume that in your original analyses (with all the temporal conditions) you included all of the 51 'effect of rest x Time#" effects in the list of confounding effects during denoising (this is the proper/default thing to do for sliding-window temporal analyses), while in the latter analyses you did not (only a single 'effect of rest' effect was included)? If that is the case, then, depending on the window length and overlap, the former analyses would remove a lot of the low frequency BOLD signal variability compared to the latter analyses as part of denoising, so your resulting original "rest" connectivity is likely focusing on the contribution of higher-frequency effects while the latter "rest" connectivity analysis is focusing on a broader frequency range, which may explain the general similarities/differences between the two analyses.
Let me know if that makes sense and/or your thoughts
Alfonso
Originally posted by ys1j13:
Dear experts and Alfonso
I recently wanted to conduct temporal network analysis, so I used temporal decomposition option in SETUP.
There is a condition called S which is entire one of four fMRI session we conducted.
I divided this session into 51 windows using temporal decomposition.
After I divided the session S, there is still intact session condition S in SETUP.
So, I expected that the intact condition S allows me to perform network analysis for entire S session.
And, I performed the analysis and produced a results for entire S.
But I wanted to make sure, so I removed the temporal conditions (so that there is only condition S) and conduct additional analysis using that condition (that there is only one condition S).
And, I produced another results for entire S using the reduced condition setup.
I anticipated that the two analysis gave me identical results, but they were clearly different (although general trend was similar).
In my knowledge, except the removing session effect in Denoising step, there is no calculation step which influence the results those analyses.
Is there a reason which causes this difference?
Best regards,
YJ
I recently wanted to conduct temporal network analysis, so I used temporal decomposition option in SETUP.
There is a condition called S which is entire one of four fMRI session we conducted.
I divided this session into 51 windows using temporal decomposition.
After I divided the session S, there is still intact session condition S in SETUP.
So, I expected that the intact condition S allows me to perform network analysis for entire S session.
And, I performed the analysis and produced a results for entire S.
But I wanted to make sure, so I removed the temporal conditions (so that there is only condition S) and conduct additional analysis using that condition (that there is only one condition S).
And, I produced another results for entire S using the reduced condition setup.
I anticipated that the two analysis gave me identical results, but they were clearly different (although general trend was similar).
In my knowledge, except the removing session effect in Denoising step, there is no calculation step which influence the results those analyses.
Is there a reason which causes this difference?
Best regards,
YJ
Threaded View
| Title | Author | Date |
|---|---|---|
| ys1j13 | May 24, 2020 | |
| Alfonso Nieto-Castanon | May 27, 2020 | |