Post Your Answer

Hi Bob,   It is essential to do a stationary test before executing OLS. However, if you plan to use non-stationary variables in the regression methods, the process will give you spurious results.   Still, the problems are faced by the researchers when it comes to causality. But a unified theory will help the researchers to use statistical methods for predicting the future, and it will be helpful while experimenting with the results.

I agree with @Rohan Maggoo. We need a proper theory on using statistics in experimental studies to manage the causality, which cannot be resolved quickly.    Making the data stationery is a necessary process that you cannot avoid. If you try to use it without converting the data stationary, then spurious correlation will become a dilemma.   Please check the below link to know more about Regression models,    https://bookdown.org/rushad_16/TSA_Lectures_book/regression-models.html   Good Luck

If you plan not to make your data stationary, you can try other methods like cointegration, where you can integrate two or more non-stationary time series. In addition, you can use two or more sets of variables that help to identify long-run parameters.    Regards, Matt

I don’t know how accurate your results will be if you use the cointegration method. But if you plan to do so, you can try using Autoregressive Distributed Lags (ARDL) models. If you are going to convert your variable as stationary data, I suggest you go with a Differencing method. However, this model is outdated, and now people are using it rarely. Anyhow, non-stationary variables are not a problem, but they have their limitations and challenges.

To run your regression, you can use time-series data stationary of the same range. But you cannot change your results when it comes to dependent variables and independent variables, right? For instance, when you are planning to make your dependent variables stationary, it is essential that your independent variables also be stationary. If not, then your regression results will not be trusted. Try to analyze and pick the theory that allows you to find the other variables without disturbing the basic principles of time series analysis.   All the Best

  Hello Bob,   If your variables are non-stationary in an OLS regression will not provide you a meaningful result and probably will end up with spurious results. And it is the best choice to make your variables stationary to avoid spurious regression and unpredictable outcomes. Moreover, make your independent variables as well as dependent variables as stationary data.   Good Luck