 |
 |
Hi Donald The non-white noise (head-motion, cardiac pulsation, respiration) causing temporal correlation in fMRI residuals has spatial structure, and since a lot of these sources are oscillatory in nature they are much more complex than a simple AR(1)+ white noise model, which cannot even model a single oscillation. However a lot of these noise sources are relatively broadband in nature (e.g. the cardiac rate could easily change around +/-0.1Hz during a fMRI session) so with long TR (about 2s) this roughly corresponds to the sampled bandwidth, and thus the oscillations are hard to identify directly from the data, since things look more white than they are. In this case the AR(1) is maybe not so bad, but in a future with shorter possible TR's which should be attractive to people looking at transients in the BOLD response (e.g. DCM) the AR(1) model will start to fail dramatically! Now concerning your AR(1) estimate. The reason it changes is because the mask defined by your regressors of interest changes. The philosophy behind the pooled AR(1) estimate in SPM is that we only need to ensure i.i.d. noise where we want to make inference about our effect. My viewpoint it that the motion regressors should not contribute to this mask. Similarly in your case one could argue that it should only be the PPI regressor which defines this mask because it is within the mask defined by this regressor you are interested in making inference. I would however remove e.g. motion effects in your PPI regressor before doing this, otherwise your AR(1) estimate will once again be biased by voxels where there is only an effect of motion. If you think that your original mask is better at defining the areas of interest than your PPI regressor then you could use the old mask. Practically, as you probably know, you can define the columns of interest (iC) and nuisance (iG) in SPM.xX. Unfortunately this is not yet a part of the batch interface, but I am sure you can solve this. Best Torben Torben Ellegaard Lund Associate Professor, PhD Center of Functionally Integrative Neuroscience (CFIN) Aarhus University Aarhus University Hospital Building 10G, 5th floor, room 31 Noerrebrogade 44 8000 Aarhus C Denmark Phone: +4589494380 Fax: +4589494400 http://www.cfin.au.dk [log in to unmask] Den Uge:21 22/05/2012 kl. 22.33 skrev MCLAREN, Donald: > In doing PPI analysis, the AR(1) term is recomputed for each model (e.g. each seed region). Since the goal of AR(1) is to remove the autocorrelation from the data and its computed at the first level, does it also need to be computed for each PPI model? Since the temporal autocorrelation in the data is the same, I was thinking that one could use the AR(1) parameters/estimates from the first-level model instead of re-estimating the AR(1) term. Does anyone have any thoughts on using the AR(1) term from the first-level versus re-estimating it for each and every seed region? > > The estimates seem to change slightly between the first level and each ppi seed region. > > Best Regards, Donald McLaren > ================= > D.G. McLaren, Ph.D. > Postdoctoral Research Fellow, GRECC, Bedford VA > Research Fellow, Department of Neurology, Massachusetts General Hospital and > Harvard Medical School > Website: http://www.martinos.org/~mclaren > Office: (773) 406-2464 > ===================== > This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED > HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is > intended only for the use of the individual or entity named above. If the > reader of the e-mail is not the intended recipient or the employee or agent > responsible for delivering it to the intended recipient, you are hereby > notified that you are in possession of confidential and privileged > information. Any unauthorized use, disclosure, copying or the taking of any > action in reliance on the contents of this information is strictly > prohibited and may be unlawful. If you have received this e-mail > unintentionally, please immediately notify the sender via telephone at (773) > 406-2464 or email. >