Thanks for your article - a couple of questions. My background is that in the dim dark past I studied neuropsych and then worked as an EEG tech (mostly epilepsy screening with infrequent psych patients) for 4 years (2000-04) before going to work for the company that makes Synamps 2 :)
Now, I've only skimmed your method and results as I'm a bit flu-ridden and hence vague, but a couple of things popped out. The first is that you stated there was no control for medication. In our experience doing EEGs on folks on certain psych drugs, there was usually an increase in higher band activity (more beta, much less slow wave) - these drugs would create an unusual but not clinically abnormal EEG. It was rare to see such an EEG in someone not on those drugs.
The second is that for doing mathematical analysis on the EEG, I've had it trained out of me that resting EEG is okay to use. Some task, any task, no matter how easy, gives a more reliable baseline - since the EEG is quite dependent on arousal state, without a basic task, you don't have much control over whether the person is sitting there thinking about having a nap or highly alert and fretting over some unrelated item. For epilepsy you want them drowsy and nearly asleep as that lowers the bar for spike-and-wave activity, but unless a similar thing happens in schizophrenia and slowing, I'd think that doing a task would be superior to resting EEG.
I wonder if accounting for these issues might increase your hit rate? Like I said, I only skimmed the method and results, so I may have missed something that makes these points less relevant.
As an aside, the epileptic EEG sample looks really quaint and old-fashioned, since it has the curved needle-on-paper distortion. It doesn't need to be updated, it just looks like Ye Olde EEG to my eyes... :)