Log likelihood criterion

• Since the product of these terms can be tiny, we can effectively maximize the log of the likelihood
• This is equivalent to the previous one
• overall maxima of the two criteria must be in the same place, so the best model parameters φˆ are the same in both cases
•