Error message: *** glibc detected *** free(): invalid pointer: 0x00002aab002dadf0 **

[Robert Oostenveld]

Hi Frederic

On 6 May 2008, at 15:07, Frederic Roux wrote:
> I found out what the error was related to. It's related to the
> 'feature accel off' option.
>
> When I launch my analysis with that option the bug does not appear.
> Can you tell me why that is so?

I don't know what the relatino between "feature accel off" and the
emx file is, I have not heard about problems with that before.
However, I know that in an older matlab versino (don't recall the
exact version number) where that feature wa sintroduced, that there
were problems with persistent variables (used underneath the
cfg.feedback option in fieldtrip). To fix those matlab bugs, it was
also needed to turn accel off.

> I have 2 more questions:
>
> 1. Can I contribute to Fieldtrip with my functions? I've written a
> baseline Correction function that computes Decibeland Z-scores.
> I've also got other functions that could be useful to other users.

The different baselineling options would be interesting to
incorporate into the existing freqbaseline function. Would that be
doable? If so, then they would automatically be available for
plotting (cfg.baselinetype refers to freqbaseline).

For the other functions: In the past other people within the Donders
have contributed some miscellaneous functions as well. I have added
them, but over the years (as they have not been maintained) the
functions might have lost their usefullness. Those functions sofar
also have not all been included in the external ftp release version
of fieldtrip. However, I certainly would like to encourage people to
contribute to fieldtrip. If the functions are sufficiently usefull
and fieldtrip-like (regarding their interface and documentation) then
I am also happy to add them to fieldtrip-main and play an active role
in maintaining them. And for small-helper  functions I could add a
directory fieldtrip/contrib where those functions could go. An
example functions that I think would be better of in fieldtrip/
contrib would be source2sparce.m (which is a helper function taht I
made myself, but it is not very fieldtrip-like).


> 2. When I compute a Wavelet transformation of my MEG signals I get
> an aquward looking power spectrum that looks like the signal is
> full of epileptic spikes.
>  It kinda looks like a 'flame pattern'. I remember that you
> mentioned it during the Fieldtrip toolkit but I can't remember what
> it was related to. Can you tell me what this means?

There are probably short and sharp transients in your data (i.e.
"blips"). The morlet wavelets for the higher frequencies get shorter
and shorter, and increase the frequency bandwith at trhe same time.
So a broad-looking blip at a low frequency becomes smaller at higher
frequencies, hence the flame pattern.

> The funny thing is that this does not happen when I do a multitaper
> analysis of the same signal.

With multitapers you probably use a longer t_ftimwin for the high
frequencies (where the morlet wavelets are short, especially with
your cfg.width=4). The increased timewindow together with the
multiple tapers results in a more robust power estimate, which is
less affected by noise (your blips).

best regards
Robert

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/fcdonders/fieldtrip.