Ordinary/Conditional Least Squares vs. Maximum Likelihood


New Member
Can anyone explain in layman's terms when we should use maximum likelihood as opposed to ordinary or conditional least squares for regression or time series analysis?

I know that when the regressors are stochastic we should use ML because the unbiased assumption of OLS/CLS will be violated, but there has to be other reasons as well. Like what statistical properties of the regressors or data set would make you prefer ML to OLS/CLS?