How can one ever run OLS on real world data


Fortran must die
I think it is generally agreed 1) OLS is commonly used including in the academic literature and 2) it is invalid to use it in its standard form when the data is non-stationary. I think there would also probably be agreement with the following "Nelson and Plosser (1982) argue that almost all macroeconomic time series onetypically uses have a unit root. The presence or absence of unit roots helps to identify some features of the underlying data generating process of a series."

Whether that is true for non-economic data I have not seen comments on, but my guess is that it also is commonly (almost always) non-stationary over several years of data.

So how do we use OLS, and similar General Linear methods which have the same issues. How can they be valid.

Obviously a lot of statistical analyst don't think this is an issue :) So why don't they, why doesn't non-stationarity negatively influence these very common approaches?