help
help > RE: interpreting results with and without within-subject normalization (v2v)
Jul 29, 2021 11:07 PM | Alfonso Nieto-Castanon - Boston University
RE: interpreting results with and without within-subject normalization (v2v)
Hi Emily,
Right, that is a good question, in general the normalization will remove/control-for potential global differences in these measures across subjects and help you focus on localized effects. That can be helpful in cases where those global differences are mainly either noise or caused by unrelated sources of variability across subjects (but of course it can also be hurtful in cases where those global differences are either meaningful or driven by the same effects that you are trying to measure). In your case, if the global normalization procedure appears to show similar while larger/stronger effects, then that would be consistent with the former scenario (and of course depending on the choice of voxel-to-voxel measure global differences in these measures across subjects may be driven by different factors; e.g. global differences in local correlation measures may indicate residual motion-related effects that are sometimes hard to correct by standard denoising strategies)
Hope this helps
Alfonso
Originally posted by Emily Stern:
Right, that is a good question, in general the normalization will remove/control-for potential global differences in these measures across subjects and help you focus on localized effects. That can be helpful in cases where those global differences are mainly either noise or caused by unrelated sources of variability across subjects (but of course it can also be hurtful in cases where those global differences are either meaningful or driven by the same effects that you are trying to measure). In your case, if the global normalization procedure appears to show similar while larger/stronger effects, then that would be consistent with the former scenario (and of course depending on the choice of voxel-to-voxel measure global differences in these measures across subjects may be driven by different factors; e.g. global differences in local correlation measures may indicate residual motion-related effects that are sometimes hard to correct by standard denoising strategies)
Hope this helps
Alfonso
Originally posted by Emily Stern:
Hi All,
I ran two voxel-to-voxel analyses - one with the "normalization" box checked off (for within-subject spatial normalization) at the first level and one with it not checked off. The results are slightly different in terms of the threshold required to see an effect (within-subject normalization at the first level produced findings at a more stringent threshold than analyses without normalization). I'm trying to figure out why. Is this simply a case of increased signal-to-noise when using normalization, or could there be other relevant factors?
Thanks for any insight,
Emily
I ran two voxel-to-voxel analyses - one with the "normalization" box checked off (for within-subject spatial normalization) at the first level and one with it not checked off. The results are slightly different in terms of the threshold required to see an effect (within-subject normalization at the first level produced findings at a more stringent threshold than analyses without normalization). I'm trying to figure out why. Is this simply a case of increased signal-to-noise when using normalization, or could there be other relevant factors?
Thanks for any insight,
Emily
Threaded View
Title | Author | Date |
---|---|---|
Emily Stern | Jul 29, 2021 | |
Alfonso Nieto-Castanon | Jul 29, 2021 | |
Emily Stern | Jul 30, 2021 | |
Emily Stern | Jul 29, 2021 | |