Is there a reason for the following text regarding the --meas argument of
mne_do_forward_solution:
"It is recommended that the average file is employed for evoked-response
data and the original raw data file otherwise."
If all the measurement file contributes is the sensor location/orientation
info and the device-head transform, does it matter from which file - raw or
averaged - these data are pulled? The quoted text from the manual would
suggest that a new forward solution be computed for each average file even
if they all derive from the same raw measurement, but that doesn't seem to
make sense.
My reading of that is that you should use an average file for --meas
if you are processing averaged data, and a raw file if you are
processing raw data. I do not believe that there are any issues with
having multiple average files from the same raw data file and only
using one forward solution calculated from one of the average files.
Just that it is not recommended that you use the raw file to generate
the forward solution.
to add to this, as I understood it, the reason for this is that channels selected are taken into account by the leadfield computation. In case, raw and average include different channels (as often is the case) you might get slightly different results.
Anyone correct me if I'm mistaken.
Cheers,
Denis
HI Andrew,
My reading of that is that you should use an average file for --meas
if you are processing averaged data, and a raw file if you are
processing raw data. I do not believe that there are any issues with
having multiple average files from the same raw data file and only
using one forward solution calculated from one of the average files.
Just that it is not recommended that you use the raw file to generate
the forward solution.
Hi Andy,
I am guessing that is there just to remind us that we could have
applied transformations to the head coordinate frame like when using
'maxfilter -trans ...' and hence it is best to use the --meas file to
which the inverse generated from this particular forward would be
applied.
unless something happened to the raw file since the evoked
was computed the measurement information (channel
location, head transformation etc.) should be identical between the raw
and the evoked fif file. I guess this recommendation is mostly
to avoid mistakes when you work on a particular evoked dataset.
So, unless some change in the coordinate transforms has occurred between
the raw and evoked files, there should be no difference when using
either for computing the forward solution, correct?
So, unless some change in the coordinate transforms has occurred between
the raw and evoked files, there should be no difference when using
either for computing the forward solution, correct?
This is correct. I have tried to find the rationale for the sentence in the Manual. I think I had the idea that the automatically created forward solution file name (see --fwd option) would better reflect the associated data if the convention is followed. However, I might do it differently now.