Normalizig Data

[Bezi Nyem]

Hi everyone,

I am trying to normalize my data to run panel regression. So far I have tried to:

1. transform the data to log form

2. normalize data as a panel country using the following commands
egen min_x1=min(x1)
egen max_x1=max(x1)
gen normalx1= (x1-min_x1)/(max_x1-min_x1)

3. normalize data by country and by variable using
su x1, meanonly
gen normal_x1=(x1-r(min))/(r(max)-r(min))

3. smoothing the data using tssmooth and also HP filter

However my data for all variablea is still skewed no matter what i tried. I have checked it using histogram frequency graphs, density plots etc.

I would appreciate some help on this.

Thank you

[Carlo Lazzaro]

Bezi:
I'm not clear with what you're after.
Why "normalizing" data to perform a panel data regression?

[Bezi Nyem]

Dear Carlo,

I have provided two examples of how most of my data looks when I check for normality of data (please see poverty and gdp graphs). My problem is that no matter what I do to the data (ie whether I keep it in its original (raw) form, transform to log, use sqrt, or standardizing the data etc), I can't seem to correct the problems below (i get the same distribution with no changes whatsoever) I had hoped that by employing one of the methods, I would be able to normalize my data. I would appreciate your help.

[Carlo Lazzaro]

Bezi:
assuming that you're going to use -xtreg-,it is not mandatory to normalize variables (as it is not a prerequisite for regression procedures)..
If you are concerned about heteroskedasticity and/or autocorrelation in residual distribution, simply -cluster()- the standard errors of your predictors.

[Bezi Nyem]

Thank you Carlo