Dear researchers,

I want to demean my data to mitigate cross-sectional dependency. I read some papers that demean the data first and then take log. Could anyone kindly enlighten me on how to demean data?

If the data are demeaned as below:

eitdemean = eit −ei ̄

Then if I take log of the demeaned data, many values will be missing, as many demeaned data are negative. I know some people add a certain value to the demeaned data to make them positive, but if I add the same number to all the variables, the coefficients between the variables will change. So could anyone give me some advice on how to demean the data?

Many thanks!