Back to PLS HelpPosted on 02/20/18 16:53:34
Number of posts: 3
Hello everyone,
I have used PLS before but it has been quite a while since I have done it. My question pertains to onset specification to enter into PLS (in TR).
To make things simple, imagine the onset for condition A occurs at 2.5 seconds. My TR is 1067. Thus my onset time in "TRs" is 2.34 (i.e., 2.5/1.067), if my math is correct. However, I was noticing that the onset is "zero indexed" for PLS. I am confused what the onset would be in TR then....Does this mean it is adjusted to 1.34? Help!
A second (and third) question pertains to preprocessing: Does it matter what order preprocessing occurs. Data were preprocessed in FSL in which smoothing occurs before MNI transformation. Hence smoothing, then MNI transformation. Is this OK? Finally, does PLS require temporal filtering? My recollection is that it does not.
Many thanks and sorry for the basic questions,
DJ
Replies:
Posted on 02/21/18 05:29:18
Number of posts: 394
Hello everyone,
I have used PLS before but it has been quite a while since I have done it. My question pertains to onset specification to enter into PLS (in TR).
To make things simple, imagine the onset for condition A occurs at 2.5 seconds. My TR is 1067. Thus my onset time in "TRs" is 2.34 (i.e., 2.5/1.067), if my math is correct. However, I was noticing that the onset is "zero indexed" for PLS. I am confused what the onset would be in TR then....Does this mean it is adjusted to 1.34? Help!
A second (and third) question pertains to preprocessing: Does it matter what order preprocessing occurs. Data were preprocessed in FSL in which smoothing occurs before MNI transformation. Hence smoothing, then MNI transformation. Is this OK? Finally, does PLS require temporal filtering? My recollection is that it does not.
Many thanks and sorry for the basic questions,
DJ
Hi DJ
Regarding the onset issue, the first volume is consider t=0 so you should adjusts your onsets accordingly as you suggest.
As for preprocessing, to conditions are basiclaly the same as for any other analyses, so keep the same order as you would for any complementary analyses you might be doing, including using temporal filtering. It does matter less for event-related fMRI in PLS since we look at local deviations around event onsets, but again to compare with other analyses you will want to keep the same processing steps.
cheers
Randy
Posted on 02/21/18 14:43:59
Number of posts: 3
Hi Randy,
Thanks for your prompt response! :) I was able to input the data into PLS beautifully and am following the manual from the Grady lab. Everything has gone smoothly. Data were preprocessed according to standard specifications and put into the MNI template and resampled to 4mm3 prior to running PLS. My one concern is when I click on the "check image orientation" the crosshair is not at 0 0 0. Is this problematic? It seems to be so based on p.15 of the manual. If it is a problem, what can one do to fix? Data were preprocessed in FSL. Sorry for the naive question.
Many thanks,
Daniela
Posted on 02/22/18 00:25:59
Number of posts: 394
Hi Randy,
Thanks for your prompt response! :) I was able to input the data into PLS beautifully and am following the manual from the Grady lab. Everything has gone smoothly. Data were preprocessed according to standard specifications and put into the MNI template and resampled to 4mm3 prior to running PLS. My one concern is when I click on the "check image orientation" the crosshair is not at 0 0 0. Is this problematic? It seems to be so based on p.15 of the manual. If it is a problem, what can one do to fix? Data were preprocessed in FSL. Sorry for the naive question.
Many thanks,
Daniela
Difficult to know if there is a problem. Where on the image do the crosshairs go? All I can suggest is checking whether the cursor locations seem correct. So if yoy click on the anterior commisure, is it at 0,0,0? You can also check using FSLview whether the locations indicated there correspond to what you see in PLSgui
Posted on 04/27/20 14:36:57
Number of posts: 9
Hi DJ
Regarding the onset issue, the first volume is consider t=0 so you should adjusts your onsets accordingly as you suggest.
As for preprocessing, to conditions are basiclaly the same as for any other analyses, so keep the same order as you would for any complementary analyses you might be doing, including using temporal filtering. It does matter less for event-related fMRI in PLS since we look at local deviations around event onsets, but again to compare with other analyses you will want to keep the same processing steps.
cheers
Randy
Hi Randy,
Thank you for your helpful replies.
If understand correctly, your reply suggests that we should prepare onsets by doing: (Onset in sec/TR duration) - 1
However, I hear that other people have just done: Onset in sec/TR duration.
May I ask for a confirmation of what is correct?
Thank you immensely.
Annick
Posted on 04/28/20 11:13:49
Number of posts: 394
Hi Annick- the counter in PLS set the first volume of a time series to zero, so counting starts at 0
Login to reply to this topic.
