[Eeglablist] Re: using calibration pulses and establishing 2-channel references (arno)

[Clayton Hickey]

Hi Ben,

You asked 2 questions, one about how to 'normalize' data based on a  
calibration signal, and one about how to reference data to an average of  
the two mastoids when one mastoid is recorded. I've dealt with both these  
things myself (I'm also using ERPSS with SA amplifiers), so to add to  
Arno's insightful comments:

1.) You don't really want to normalize the data, I don't think, not in the  
way most people understand the term normalize. This is why Arno expresses  
a bit of puzzlement over your request. What you want to do is give a  
metric to the data, based on a calibration signal. You probably have  
something like a 20 uV cal pulse, and you want to find out how many A-to-D  
board units this is represented by. You can then apply this to the data  
that you've imported, which is probably in A/D units.

You can do this by calculating the average magnitude of the cal pulses  
across each  pulse (there will be variability both within the pulse,  
especially at the leading edge due to filter artefacts etc., and across  
pulses), divide this value by 20 (to get a 1uV = xx A/D units conversion  
rate), and apply this conversion value to your data (ie. divide A/D metric  
data by the conversion rate you created).

2.) You've recorded data in reference to, say, M2 and want to reference to  
an algebraic average of the left and right mastoids. You can do so by  
simply subtracting half of the recorded M2 signal from the data. This is  
so because:

we want: e1 - 0.5*(M1+M2)
we have: (e1 - M1) and (M2 - M1), in that each electrode is referenced to  
M1, and M2 is also recorded in reference to M1.

e1 - 0.5*(M1+M2)
= e1 - 0.5*M1 - 0.5*M2
= e1 - M1 + 0.5*M1 - 0.5*M2
= e1 - M1 - 0.5*(M1 - M2)
= (e1 - M1) [which we have] - 0.5*(M2 - M1) [which we have]

This should be pretty easy to do from the command line in matlab.

[ note: I cannot take any credit for this clever manipulation; Jon Hansen  
addresses this issue in an internal supporting document for ERPSS, and the  
equation above is taken more or less verbatim from there! ]


best,

Clayton Hickey
Simon Fraser University