PCA with multiple subject's data

Help
Isaac
2012-06-01
2013-02-26
  • Isaac
    Isaac
    2012-06-01

    Is it possible to run a PCA on multiple subjects' data (not grand average) simultaneously using the toolkit?  I would like to characterize each person's waveform by the same temporal and spatial components to see if individual differences in the component corresponding approximately to the P3 are associated with task performance.  I would prefer to use the 2-step temporospatial PCA, but need to run it on the individual-level data simultaneously rather than on the on the grand average waveform.

    Thanks!

     
  • Joe Dien
    Joe Dien
    2012-06-01

    Yes!  In fact, this is the recommended approach.  I showed in my 2010 paper that using subject averages yields better results than a grand average as the subject variance provides the PCA with additional information.  You'll want to combine the subject averages into a single file that contains the separate subject averages in it.  EGI files and .study formats can handle multiple subject averages and can be imported directly.  Other formats may require using either reading them in using "single file mode" or using the Edit function's "append subject" option to combine the subject averages.  See the tutorial for instructions on what to do.

    Cheers!

    Joe

     
  • Joe Dien
    Joe Dien
    2012-06-01

    or if you're starting with single trial data, the Average function will generate a single combined subject average file if you average them together at the same time.

     
  • Isaac
    Isaac
    2012-06-02

    Great, that makes perfect sense.  Thanks so much!