[FieldTrip] Plotting freq/chan/time data and ft_freqstatistics RAM issues
julian.wang at ucl.ac.uk
Mon Nov 21 17:54:49 CET 2011
I'm trying to plot some data that I've just finished the analysis for, and
was wondering if anyone had any advice on how it should be done. I've just
done a frequency analysis (with a freq range of 5-80Hz) and a permutation
based stats test on a set of EEG data, and it has come out with several
significant clusters, but I'm not sure how to represent this visually. I
tried using ft_clusterplot function, but it doesn't seem to like multiple
frequencies. Is there a way of making it plot only one frequency?
I was also wondering if it was possible to display the data in form of a
short movie clip, showing something like a cube, with the x,y plane of the
cube being a spatial representation/scalp map of the EEG data, and with the
z plane representing the frequency range (in my case, 5-80Hz), and having
the entire cube change with time.
This might be useful to help visualise the significant clusters, especially
if the clusters could be plotted individually. Is this something that's
possible at all, or is there a better way of displaying the data?
Another problem that I keep on running into is when running the permutation
based stats tests, I keep on getting out of memory errors. Currently I'm
running Windows 7 64bit with Matlab 2011 64bit and have 4GB of physical RAM,
but I've also assigned an additional 16GB of virtual RAM to help it along.
This is what Matlab reports:
Maximum possible array: 17666 MB (1.852e+010 bytes) *
Memory available for all arrays: 17666 MB (1.852e+010 bytes) *
Memory used by MATLAB: 509 MB (5.335e+008 bytes)
Physical Memory (RAM): 4094 MB (4.293e+009 bytes)
* Limited by System Memory (physical + swap file) available.
But even with 20GB of RAM assigned, I find that I can only do around 1000
permutations on my data set of 16 subjects, each with around 700 trials, 2
conditions, 75 frequencies, 260 time slices. Is there a way of increasing
the limit on the number of permutations that I can run (without increasing
the assigned amount of virtual RAM)? Or would it be better to just run a
loop and calculate the stats at each frequency individually?
Any help would be greatly appreciated.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the fieldtrip