Back to PLS HelpHow to freely specify a condition covring a time range
Posted on 06/29/09 08:13:24
Number of posts: 60
Hello all,
According to my knowledge about PLS, it would be a kind of impossible to determine and investigate a condition (pattern) covering a time range between two pre-specified conditions? This kind of analysis would be feasible to be done using other multivariate approach like multi-voxel pattern analysis (MVPA) due to its onset-free characteristic.
They main intention for asking this is to varify if there would be a pattern (term it as pre-identification) few seconds before conscious identification.Thereby, there is no precise information about the onset of this pre-identification but we know that there could be possibly a build-up pre-identification few second before identification and after stimulus onset.
By the firse analysis, I made an assumption that pre-ide would taken place 2TR before identification, but it seemed that PLS could not dissociate between pre-ide and ide in that case.
Additionally, it would be possible that the build-up pattern related to pre-iden take place somewhere before or after 2TR which render the assumption that we made unjustified.
Is their any kind of tricky solution to specify this condition in PLS?
/Alireza
According to my knowledge about PLS, it would be a kind of impossible to determine and investigate a condition (pattern) covering a time range between two pre-specified conditions? This kind of analysis would be feasible to be done using other multivariate approach like multi-voxel pattern analysis (MVPA) due to its onset-free characteristic.
They main intention for asking this is to varify if there would be a pattern (term it as pre-identification) few seconds before conscious identification.Thereby, there is no precise information about the onset of this pre-identification but we know that there could be possibly a build-up pre-identification few second before identification and after stimulus onset.
By the firse analysis, I made an assumption that pre-ide would taken place 2TR before identification, but it seemed that PLS could not dissociate between pre-ide and ide in that case.
Additionally, it would be possible that the build-up pattern related to pre-iden take place somewhere before or after 2TR which render the assumption that we made unjustified.
Is their any kind of tricky solution to specify this condition in PLS?
/Alireza
Replies:
Untitled Post
I'm Online
Posted on 06/29/09 08:47:13
Number of posts: 229
By
the firse analysis, I made an assumption that pre-ide would taken place
2TR before identification, but it seemed that PLS could not dissociate
between pre-ide and ide in that case.
Additionally, it would be possible that the build-up pattern related to pre-iden take place somewhere before or after 2TR which render the assumption that we made unjustified.
One option would be to analyze a long time window that includes both 'pre-event' activity as well as 'post-event' activity.
Use the TR that corresponds to event onset to normalize (zero) the data - this will allow you to determine if post-event and/or pre-event activity is most strongly related to your task manipulation.
nancy
Additionally, it would be possible that the build-up pattern related to pre-iden take place somewhere before or after 2TR which render the assumption that we made unjustified.
One option would be to analyze a long time window that includes both 'pre-event' activity as well as 'post-event' activity.
Use the TR that corresponds to event onset to normalize (zero) the data - this will allow you to determine if post-event and/or pre-event activity is most strongly related to your task manipulation.
nancy
Untitled Post
Posted on 06/29/09 11:16:28
Number of posts: 60
Thank you Nancy for your quick reply!
let's see if I properely understand your second suggestion.
let's say we have onset for identification and stimuli onset as follow:
Run1Identification: [27,49,73,186,324,351,395]
Run1stimulus onset: [14, 39, 60,113,151,175,211,318,336,362,386]
Do you mean that I have to specify 27 for the relative ref.scan onset for identification and 14 for relative ref.scan onset for stimulus onsrt? If so, according to the code (fmri_get-datamat) the baseline-start would be [27,49,73,186,324,351,395] and the baseline-end would be [27,49,73,186,324,351,395]+27 for identification.right?
If we have two run with different onset, which one should be taken as ref.scan onset.
thanks alot for your help!
/Alireza
let's see if I properely understand your second suggestion.
let's say we have onset for identification and stimuli onset as follow:
Run1Identification: [27,49,73,186,324,351,395]
Run1stimulus onset: [14, 39, 60,113,151,175,211,318,336,362,386]
Do you mean that I have to specify 27 for the relative ref.scan onset for identification and 14 for relative ref.scan onset for stimulus onsrt? If so, according to the code (fmri_get-datamat) the baseline-start would be [27,49,73,186,324,351,395] and the baseline-end would be [27,49,73,186,324,351,395]+27 for identification.right?
If we have two run with different onset, which one should be taken as ref.scan onset.
thanks alot for your help!
/Alireza
Untitled Post
I'm Online
Posted on 06/29/09 11:35:37
Number of posts: 229
Hi Alireza...
Since I'm not familiar with your study design, those TR #s are a bit confusing to me, so let me give you a generic example.
You have an event that starts at TR#10. If you want to look at 8TR-pre- and 8TR-post-stimulus activity for in a single analysis, referenced to event onset, you would:
1) Set Event onset to TR#2
2) Set Temporal Window size to 17
3) Set Relative Reference Onset to +8 - TR#10 now becomes the zero point for all TRs
4) Set Number of Reference to 1
that should give you a zero value at TR10 (much as you get now if you use the first scan as the reference point), and deviations around that point reflecting activity preceding and after the event.
n
Since I'm not familiar with your study design, those TR #s are a bit confusing to me, so let me give you a generic example.
You have an event that starts at TR#10. If you want to look at 8TR-pre- and 8TR-post-stimulus activity for in a single analysis, referenced to event onset, you would:
1) Set Event onset to TR#2
2) Set Temporal Window size to 17
3) Set Relative Reference Onset to +8 - TR#10 now becomes the zero point for all TRs
4) Set Number of Reference to 1
that should give you a zero value at TR10 (much as you get now if you use the first scan as the reference point), and deviations around that point reflecting activity preceding and after the event.
n
Untitled Post
Posted on 06/29/09 12:50:00
Number of posts: 60
Thank a lot Nancy!
It was very handy and useful!
It was very handy and useful!
Login to reply to this topic.
