How to run PLS Analysis on PET data

- PET datamat file

To run PLS, you need to have a one or more datamats ready,
one for each separate group of subjects. All the datamats
should have the same selected conditions. Click "Add" button
to select a datamat and add it into a group. All the
selected datamats will be stacked together rowwise for
analysis. The "Number of groups:" field will display how
many groups you have added in.

If the "Save Datamats" checkbox is selected, the 2D datamat
will be saved to a cell array "newdata_lst" in the specified
PETresult.mat file. Each cell represent a group.

"Full path" toggles the full path on/off.





- Deselect Menu

Once you have added datamat into group(s), you can click
"Deselect" menu in the menu bar. There are two items: 
"Deselect Conditions" and "Modify Behavior Data". Click 
"Deselect Conditions" first, and select which conditions 
you want to analyze. If you are doing Behavior analysis,
you can modify the behavior data right before clicking
the "Run" button. However, you need to make sure the row
order of behavior data, which should be in the form of 
"subject-in-condition". Also, make sure you modify 
behavior data after you select the conditions.





- PLS Options

Mean-Centering PLS option produces results equivalent to
the Helmert matrix version described in Lobaugh, et al.
(2001).  The grand mean across tasks is calculated, and all
individual datapoints are subtracted from this value. The
newly generated datamat is subjected to singular value
decomposition to obtain BrainLVs (Salience), singular values,
and other useful results for further analysis.

If you run a Behavior PLS, you may also include a Posthoc
Data File. This text file contains contrasts across the
conditions and/or groups in the analysis, (one contrast per
column, one line per condition).  This will test for
specific differences in brainbehaviour correlations.

Most recently, we have developed a "non-rotated" version of
Task PLS, wherein a priori contrasts are used to restrict the
patterns derived from PLS.





- About computation parameters

Permutation tests establish the significance of the latent
variables. The bootstrap results establish the stability of
the LV saliences, and in the case of a BehaviourPLS, the
confidence intervals for the brain-behaviour correlations.

Typically, the Number of Permutation is usually from 300 to
500 (the higher the better but longer), and the Number of
bootstrap is usually set to 100. If you are running Behavior
PLS with Number of Bootstrap not equal to zero, you can also
choose the Upper Limit of the Confidence Level for the brain-
behaviour correlations. The default Confidence Level is set
to 95.