[Eeglablist] Single trial data in newtimef –follow-up questions

[Arnaud Delorme]

Dear Ido,


> 1. I understand that the single trials in the "tfdata" are not baseline corrected. How can I correct each trial relative to it`s own baseline? I`ve tried to use the "powbase" output, but if I understand correctly – this is the mean baseline of all the trials, right?


Yes, this is correct. Use the 'trialbase' option of EEGLAB to perform baseline correction at the trial level. Only do that if you know what you are doing. Some issues arise with single-trial baseline correction that we do not understand fully yet and that would be too long to discuss here.


> 2. What is the best way to calculate the time course of a certain band (e.g. the power in the 5-15 Hz band as a function of time)? One obvious way is to take the "ERSP" matrix and calculate the mean activity over that frequency range, but then I won`t have any error bars. In addition, I prefer to display the time course as % signal change instead of dB – is it possible?


Everything is possible :-). The option 'plottype' can be set to 'curve' to plot specific frequency bands. However it will not plot error bars. To display power in % of baseline use the input 'scale', 'abs' (the default 'scale', 'log' use dB as unit). 


> 3. Arnaud – in one of the previous messages you wrote that: "Do not use the log transformation (default) when you work with baseline corrected single trials or some values can blow off". How can I disable the log transformation?



Same as above, use 'scale', 'abs'.
Best,

Arno