Back to PLS Help
Raw data
Posted on 08/22/08 10:29:31
Number of posts:
Dear List
I have one more question regarding the value of the data matrix for specified coordinate[x,y,z] before SVD be applied to it.can anybody tell me how one can access this value.Does this value equal with the raw data value?if not, is that only column-wise mean value subtraction
that cause the difference?how can we access the pure value which is exactly equal to raw data value for the same coordinate.
Thanks
Replies:
Untitled Post
I'm Online
jshen
Posted on 08/22/08 10:41:26
Number of posts: 291
The data matrix before SVD is based on your datamat but not equal to your datamat. Depending on which PLS method options you choose, the processing algorithm is different. I am not able to list them here because they are too long and too complicated. You'd better to look at the open source code by yourself, if you really want to know the value.
Untitled Post
Posted on 08/22/08 11:23:29
Number of posts:
Hello again Jimmy,
Thank you for your prompt resonse.so Can you tell me what isdatamat_files_timestamp(how one can read it in matlab)? Isn't it the data matrix before going through the SVD?
if Yes,Are normalization with refrence scan and mean centring the only two parameters that makes this matrix different from raw data matrix?
Thanks in advance
Untitled Post
I'm Online
nlobaugh
Posted on 08/22/08 11:31:30
Number of posts: 229
David...
you can easily extract values for specific voxels using the multiple-voxel-extraction procedure. The data values will reflect your choices in the datamat creation steps (e.g., reference scan, # reference scans, normalizing, etc).
nancy
Untitled Post
I'm Online
nlobaugh
Posted on 08/22/08 11:32:46
Number of posts: 229
datamat_files_timestamp is used for housekeeping purposes to be sure that your data have not been modified before displaying PLSresults.
Untitled Post
I'm Online
jshen
Posted on 08/22/08 11:53:00
Number of posts: 291
David,
Your last question is: Are normalization with refrence scan and mean centring the only two
parameters that makes this matrix different from raw data matrix?
My answer is: First we need to clarify the definition of raw data matrix. We usually refer the values in the datamat as row data matrix, instead of the value in the raw image scan. Second, normalization with refrence scan is done during datamat creation, so this parameter is already included in the "raw data matrix". Third, as I said, depending on which PLS method options you choose, we have different algorithms, and "mean centring" is one of the methods. You may need to adjust my answer slightly if you believe that the scanned images are the "raw data matrix".
Your second question is: Isn't it the data matrix before going through the SVD?
My answer is: No. I believe I have discussed this question.
Your first question is: Can you tell me what is datamat_files_timestamp (how one can read it in matlab)?
My answer is: It is a variable in my code to check the timestamp of your *datamat.mat file, and will then compare it with the timestamp of your *session.mat file. For MRI study, because of historical reason, the session file is not
embeded into the datamat file after the datamat file is created. We have seen a lot of situation that people modified the session file after the datamat is created, but did not re-generate the datamat file. Therefore, they will encounter problem when they run PLS analysis with unmatched session/datamat files. With this variable and checkpoint, the program will nicely prompt you if you forgot to re-generate datamat after session file is modified.
Untitled Post
Posted on 08/25/08 05:51:11
Number of posts:
Hello Jimmy,
While I say Raw data,I mean Raw image data.So in that case,how one can access the raw image data matrix using PLS?( I use Task PLS for E:R.FMRI)
Secondly, while I extract multiple voxel extraction,I see 24 column which stands for 3 Voxel*8 lags and 60 rows which stands for 14 subjects*5 Condition.So what does happen to the number of trails?This matrix seems to be very small since it does not include the trails?do you make any kind of average for the trails?
Thanks,
D
Untitled Post
I'm Online
nlobaugh
Posted on 08/25/08 08:28:22
Number of posts: 229
While I say Raw data,I mean Raw image data.So
in that case,how one can access the raw image data matrix using PLS?( I
use Task PLS for E:R.FMRI)
Since the analysis is rarely (if ever) done on the 'raw' image data, we do not provide a protocol for extracting this information. If you have chosen to not use any of the normalization options when creating the datamats, your datamats would consist of the mean signal across onsets (trials) and/or runs
Untitled Post
I'm Online
jshen
Posted on 08/25/08 10:31:05
Number of posts: 291
Nancy has already replied your first question, and also replied your second question (for your previous topic).
Your third question is: This matrix seems to be very small since it does not include the trails?
My answer is the same as Nancy's for your second question.
Your fourth question is: do you make any kind of average for the trails?
My answer is Yes. In the session profile window, there is something called "merge data across all runs" and "merge data within each run". The "merge" here is actually doing average.
Login to reply to this topic.