economicsstudent wrote:For testing efficiency. And when I say the Box Jenk, I mean the BJIDENT procedure in RATS.
What's your null hypothesis and what's your alternative? If you know what those are, and compare them with what those tests are doing, then you will have much of your answer. For instance, if your null is that log X follows a random walk with increments of common variance, then the simplest form of the D-F test does have that as the null. However, the alternative in the D-F test is that log X is a stationary autoregression. That's
never the alternative in the EMH. And log X having common variance increments may be stronger than you should be assuming. So the null is wrong and the alternative is wrong, so it's probably not a good choice for what you're trying to test.