I am trying to apply OLS for time-series data that is clearly from a neither independent nor identically distributed process. The observations are hourly power system load values and the covariates include hourly temperature and historical power system load values.
Would it be reasonable to assume that hourly power system load and temperatures values are from a strong-mixing process?
By definition, it is impossible to statistically test strong mixing. However, in the event that Lo's modified R/S test points to short-term dependence in the data, would it make sense to use such a test result to corroborate the assumption of strong mixing? Is there any (other) test that can be used to corroborate the assumption of strong mixing?
