Hello,
I am writing a thesis paper on the relationship between Income Inequality and Economic Growth. In order to transform the time series into stationary, I had to take the first difference and even the second for some variables.
As I have one variable in its second difference, do I have to take the second difference of all the variables to do the regression? Is it possible to run a regression with some variables being at level 1 and some at level 2? I heard that a solution is to take the 2nd difference of all the variables in the regression, but then it is said that this is not supported that much in papers. What should I do?
Hi Kostas, In my opinion, you can fit your model with some variables which are differences, but you should check if it's necessary use it. Usually in economical data almost all variables have positive trend but it's important to determine number of differencing for each variable. It suggest using ADF test to check stationary after each differencing.
Well I thought stock price series tend to suggest negative trends are possible. I got this after study at Trump University when I collected my own data. Best wishes, David Booth
It is possible, but I think it causes some substantial problems like not meeting the presumptions of the regression method (e.g. lack of linear relationship), non-interpretability, and so on. I believe stationery damages the principles of the regression model.
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series