Back to PLS HelpUntitled Post
Posted on 04/16/09 06:36:13
Number of posts: 60
Dear all,
I am wondering whether PLS would be able to dissociate between two conditions of interest when those conditions only differ in 2 scans for all corresponding onsets? let assume for condition1 we have onset like [t1,t2,t3,....] in scans while for condition 2 we have [t1,t2,t3,....]-2 (again in scan). would PLS be able to dissociate this two conditions (lets say for TR=3) ?
Thanks in advance.
/Alireza
Replies:
Untitled Post
Posted on 04/16/09 08:07:39
Number of posts: 394
quote:
Dear all,
I am wondering whether PLS would be able to dissociate between two conditions of interest when those conditions only differ in 2 scans for all corresponding onsets? let assume for condition1 we have onset like [t1,t2,t3,....] in scans while for condition 2 we have [t1,t2,t3,....]-2 (again in scan). would PLS be able to dissociate this two conditions (lets say for TR=3) ?
Thanks in advance.
/Alireza
That's a rather difficult question to answer definitively, but so long as the signal differences are reliable, PLS should be able to detect this. Perhaps there is more to your question, however, so feel free to provide a bit more information.Dear all,
I am wondering whether PLS would be able to dissociate between two conditions of interest when those conditions only differ in 2 scans for all corresponding onsets? let assume for condition1 we have onset like [t1,t2,t3,....] in scans while for condition 2 we have [t1,t2,t3,....]-2 (again in scan). would PLS be able to dissociate this two conditions (lets say for TR=3) ?
Thanks in advance.
/Alireza
Untitled Post
Posted on 04/17/09 06:08:32
Number of posts: 60
Dear Randy,
I am trying to varify if there would be a pattern (term it as pre-identification) few seconds before conscious identification?
The stimuli (animal sounds) remained constant throughout the trial with a loop of 4 or 8 s and stimulus onset asynchrony of 60s. Subjects were instructed to fixate on a black crosshair listening to the noisy sound. Every 4 s the noise level was reduced one step until the subject press the button indicating animal identification.Then, subjects were required to continue attending to the animal sound during the remainder of the trial.during this period, a change of background prompted the subject to press the button once more (control motor response). In the described procedure with 1 target identification , 2 motor responses, 1 onset of stimuli and 1 pre-identification (we set onsets for this condition 2 TR (TR=3s) before identification) we would like to make a comparison specifically between identification and pre-identification.
Lv1 indicates onset versus identification & pre-identification (it nicely shows the build up effect in the negative direction) whereas the motor response does not contribure reliably to the pattern (please note that how much identification and pre-identification show tendancy toward each other). Lv2 reflects identification & motor versus pre-identification while the onset of the stimuli does not contribute to the pattern.This LV delineates some kind of dissociation between identification and pre-identification that I am looking for.
Now, the problem is that the (response function plot) most reliable regions from LV2 do not show dissociation (but rather similarity) between identification and pre-identification as the mean brain score plot for this LV does.This problem beside the similarity of identification and pre-identification in LV1 (in negative direction) raise the warning signal for me that perhaps PLS would not be able to dissociate between these two conditions due to some similarity in their onset time but with 2 TR differencies.
you can access the plot for LV1 and LV2 and one of the most reliable region by :
ftp://ftp.physiol.umu.se/fysiologi/out/Alireza
/Alireza
I am trying to varify if there would be a pattern (term it as pre-identification) few seconds before conscious identification?
The stimuli (animal sounds) remained constant throughout the trial with a loop of 4 or 8 s and stimulus onset asynchrony of 60s. Subjects were instructed to fixate on a black crosshair listening to the noisy sound. Every 4 s the noise level was reduced one step until the subject press the button indicating animal identification.Then, subjects were required to continue attending to the animal sound during the remainder of the trial.during this period, a change of background prompted the subject to press the button once more (control motor response). In the described procedure with 1 target identification , 2 motor responses, 1 onset of stimuli and 1 pre-identification (we set onsets for this condition 2 TR (TR=3s) before identification) we would like to make a comparison specifically between identification and pre-identification.
Lv1 indicates onset versus identification & pre-identification (it nicely shows the build up effect in the negative direction) whereas the motor response does not contribure reliably to the pattern (please note that how much identification and pre-identification show tendancy toward each other). Lv2 reflects identification & motor versus pre-identification while the onset of the stimuli does not contribute to the pattern.This LV delineates some kind of dissociation between identification and pre-identification that I am looking for.
Now, the problem is that the (response function plot) most reliable regions from LV2 do not show dissociation (but rather similarity) between identification and pre-identification as the mean brain score plot for this LV does.This problem beside the similarity of identification and pre-identification in LV1 (in negative direction) raise the warning signal for me that perhaps PLS would not be able to dissociate between these two conditions due to some similarity in their onset time but with 2 TR differencies.
you can access the plot for LV1 and LV2 and one of the most reliable region by :
ftp://ftp.physiol.umu.se/fysiologi/out/Alireza
/Alireza
Untitled Post
Posted on 04/17/09 07:57:42
Number of posts: 394
Hi Alireza,
The data you posted look pretty good. The one thing that is hard to know from the plots is what lags the effects in LV2 are maximally expressed and what the exact contrast is. One thing that would be helpful is to see the task pls brain scores plot with confidence intervals and the bootstrap ratio image. What I see in the temporal scores plot is that ident & preid differentiate early and stay as such, but the motor vs preid also contribute, which may be the stronger effect. For the HRF plot from the strongest region, is the BSR weight positive or negative?
I assume this is the regular mean-centred PLS that you ran, but if you have specific contrasts of interest, I might suggest supplementing the present analysis with a non-rotated analysis to focus on the effects you are most interested in.
Randy
The data you posted look pretty good. The one thing that is hard to know from the plots is what lags the effects in LV2 are maximally expressed and what the exact contrast is. One thing that would be helpful is to see the task pls brain scores plot with confidence intervals and the bootstrap ratio image. What I see in the temporal scores plot is that ident & preid differentiate early and stay as such, but the motor vs preid also contribute, which may be the stronger effect. For the HRF plot from the strongest region, is the BSR weight positive or negative?
I assume this is the regular mean-centred PLS that you ran, but if you have specific contrasts of interest, I might suggest supplementing the present analysis with a non-rotated analysis to focus on the effects you are most interested in.
Randy
Untitled Post
Posted on 04/17/09 10:32:55
Number of posts: 60
Hello Randy,
I am agree with you that it is difficult to make a strong claim about the lag in which the main effects in LV2 is maximally expressed but taking a quick look at BSR image shows the positive effect is mostly expressed at lag3 and negative effect is mostly expressed at lag 5&6 which surprise me abit why the effect of pre-identification(green line) is so late and why the peak for pre-identification is 2TR after the identification while the setting for the onset was opposite (reminding that the onset of pre-identi was 2TR before identifi).One the other hand, I am not agree that the motor and onset of the stimuli would contribute to the patten in LV1 and LV2 consequently, since in LV1 confidence intervals(error bars in task PLS with CI) for motor crossing zero and I would interpret this as unreliable contribution (do you agree?).This is the same for onset of stimuli in LV2.please see additional plots at: ftp://ftp.physiol.umu.se/fysiologi/out/Alireza/
Regarding your question for the HRF plot, the BSR weight is negative and that's why I expect the blue curve to be on the negative side but obviously, it is not.
Regarding your suggestion for non-rotated PLS, I have done it for identification versus pre-identifuication and it turns out that this pattern is significant, but still the problem regarding HRF plot of the reliable regions exist meaning that the plot for identification and pre-identification are mostly getting together on one side rather than keeping out of each other as it set by the contrast ([1 0 -1 0]).
thanks for your suggestions.
/Alireza
I am agree with you that it is difficult to make a strong claim about the lag in which the main effects in LV2 is maximally expressed but taking a quick look at BSR image shows the positive effect is mostly expressed at lag3 and negative effect is mostly expressed at lag 5&6 which surprise me abit why the effect of pre-identification(green line) is so late and why the peak for pre-identification is 2TR after the identification while the setting for the onset was opposite (reminding that the onset of pre-identi was 2TR before identifi).One the other hand, I am not agree that the motor and onset of the stimuli would contribute to the patten in LV1 and LV2 consequently, since in LV1 confidence intervals(error bars in task PLS with CI) for motor crossing zero and I would interpret this as unreliable contribution (do you agree?).This is the same for onset of stimuli in LV2.please see additional plots at: ftp://ftp.physiol.umu.se/fysiologi/out/Alireza/
Regarding your question for the HRF plot, the BSR weight is negative and that's why I expect the blue curve to be on the negative side but obviously, it is not.
Regarding your suggestion for non-rotated PLS, I have done it for identification versus pre-identifuication and it turns out that this pattern is significant, but still the problem regarding HRF plot of the reliable regions exist meaning that the plot for identification and pre-identification are mostly getting together on one side rather than keeping out of each other as it set by the contrast ([1 0 -1 0]).
thanks for your suggestions.
/Alireza
Login to reply to this topic.
