stationarity test and first differences with logharitms

[araihc85]

Hello!
I'm doing a regression with a semi-log model where the dependent variable is the growth of the Gross Domestic Products and the indipendent variables are the log of different variables, such as invstments, population, education rate, ecc...I want to test all variables for stationarity, so I did an Augmented Dickey Fuller test on the variable in log (ecxpect from th gdp). Is it correct? Or do I have to do the test with the variables in level (instead of in log), even if in my original regression they are in log?
Then, I found out that almost all variables are non stationary, so I estimated the first differences. Since I did the ADF test on the variables in log, I estimated the first differences using variables in log as well. Is it correct?
Maybe I didn't understand the meaning of the log very well. Anyway, is there someone that is so kind to help me??
thanks

[moojf]

I think you should use log specification for all variables to unify your variables criteria