Index is a solution? or actually no problem?

[melike.k.]

Hi,

I have just started to work with my data. I am working on 4 time series data. I will try to make a VAR analysis. All the variables are stationary at level. Here is the my question:

One variable - let's say Y - is between the range of 600-900 and other variables is varying between 0 and 5. I am curious about that maybe even my high valued variable is significant, I may loose Y's precise effects on the other variables. Am I in the right way? If so, what do you offer me? Do you think that to make an index for Y (by taking one year as a base year) can be a solution? Or should I take its logaritm? (But i think that taking logaritm will increase their number of integrated order, unnecessarily.)

Thank you very much beforehand

[trubador]

Logarithmic transformation changes the interpretation of coefficients. If you are only worried about the precision of the estimated coefficients, then you can simply adjust for the very high values by rescaling the data. For instance, you can use Y/100 as the dependent variable instead...