Spike II system is pretty nice but most of the time people use customized system for the acquisition and processing. As for depth of anesthesia, it is rather a combination of read-out that will help you. As an example, alpha-cholarlose anesthesia is considered as dissociative for long anesthesia. It kept the brain electrical activity quite high with spontaneous epileptical events, but the animal is well anesthetized if we consider the heart rate and the blood pressure...

I presented work I have done as part of my PhD project in the free paper session of  the Association of Paediatric Anaesthetists of Great Britain and Ireland (APAGBI) in May this year. I used needle electrodes and an amplitude integrated EEG monitor in neonatal rats inhaling low concentrations of sevoflurane. I showed that it can be used to demonstrate a dose-response relationship for low concentrations - up to 1.0%.

I have attached the submitted abstract and a copy of the presentation I gave. I hope it and the references contained are helpful. 

I also have a paper published in Anesthesiology reporting a method to measure a concentration-response relationship between inhaled anaesthetics and prevention of cold-stimulated vocalisation in neonatal rats. While the evidence suggests this works in isolated neonatal rats preventing ultrasonic vocalisation, I have wondered if a relationship where vocalisation of more mature rats (audible and/or ultrasound) could also be demonstrated?

Don't forget the standard clinical indicators of anesthesia depth in animals, HR, movement, hypothermia (due to loss of peripheral vasoconstriction), spontaneous respirations. It can be difficult to get just the right anesthetic depth for e.g. TcMEP and often a balance of drugs are needed such as ketamine, fentanyl, propofol.

You can calculate the EEG power as a parameter. The total power would be lowered  in rats dependent on the depth and duration of anesthesia. And the gravity frequcenty would be another parameter (be lowered too).