CCNCS Seminar Details
When are two brainwaves different? Statistical inference in electrophysiological neuroscience using Monte Carlo resampling.
|Speaker:||Professor Howard Bowman|
|Date/Time:||Wednesday 8 February 2012, 3.00pm|
Noninvasive EEG neuroscience considers electrical signals (colloquially called brainwaves) recorded at electrodes placed on the scalp. The central technical challenge is identifying signal from noise. Typically, we seek to determine the brain's electrical response to a stimulus (the signal) from amongst electrical change arising from a host of other sources (noise). The standard method to increase the signal to noise ratio is to repeat stimulus presentation and average together the resulting electrical responses. Nonetheless, determining that a difference in the electrical response to two conditions is empirically significant and not just due to random noise can be statistically hard. We present the standard approach in this context, which we argue often fails to be statistically robust. In particular, treatment of the multiple comparisons problem is often ad hoc and can be unconvincing. We discuss a set of rather straightforward methods to alleviate these difficulties using Monte Carlo resampling, the logic of maximal statistics and a Fisher scoring procedure.