Dear Ming,


The mean of the time series at each voxel is soaked up by the last regressor in the design matrix, the column of 1's.


All the best, W


From: Ming Teng <[log in to unmask]>
Sent: 06 October 2015 03:05
To: Penny, William
Cc: [log in to unmask]
Subject: Re: [SPM] A question on slice or subvolume
 
Dear Will,

Thanks for the answer. I also have a question on high-pass filter and would greatly appreciate if you or somebody else can answer that. 

When analyzing the face-repetition dataset in your 2005 paper, it's said that "each time series was then high-pass filtered using a set of discrete cosine basis functions with a filter cut-off of 128s". In my understanding, high pass filter will remove the mean of the time series as well, resulting a mean 0 time series. But when I try this in SPM, only low-frequency signal drift is removed, the mean still remains. Would you mind telling me if there's any step that remove the mean of the signal as well, or if there's additional hyper-parameters introduced that estimates the mean of the time series as well?

Many thanks,
Ming

On Fri, Oct 2, 2015 at 3:48 AM, Penny, William <[log in to unmask]> wrote:

Dear Ming,


Ideally, one would wish to fit Bayesian models with spatial priors to the whole brain as a single 3D volume. But this is too computationally

expensive. So two workarounds have been implemented.


1. Slice refers to a single two-dimensional slice through the brain i.e. with a fixed z co-ordinate. Models are fitted independently for each slice (this is of course a big step up from the usual mass-univariate approach where models are fitted independently for each voxel)


2. Subvolume refers to a set of contiguous voxels in 3D brain space that has been identified using a graph partitioning algorithm. You can read more about this option in [1]. From the Abstract: "While fMRI data are collected in slices, the functional structures exhibiting spatial coherence and continuity are generally three-dimensional, calling for a more informed partition. Models are fitted independently for each subvolume. 



(2) is more computationally intensive than (1) but will likely give better results.

Best,

Will.

[1] L M Harrison, W Penny, G Flandin, C C Ruff, N Weiskopf, and K J Friston. Graph-partitioned spatial priors for functional magnetic resonance images. . Neuroimage, 43(4):694-707, 2008.
From: SPM (Statistical Parametric Mapping) <[log in to unmask]> on behalf of Ming Teng <[log in to unmask]>
Sent: 30 September 2015 18:56
To: [log in to unmask]
Subject: [SPM] A question on slice or subvolume
 
Hi All, 

When I'm doing Bayesian 1st level analysis of face-repetition data, I have to select whether it's a "slice" or "subvolume" for the "block type" option. Could anyone tell me the difference between the two, and which option shall I select for this particular dataset? 

Thanks in advance,
Ming