Normality test

[Hester vanderGaag]

I have a panel data set. In my regression model I have GDP per capita, GDP per capita squared, labour productivity and an interaction term of GDP per capita and labour productivity. Should I only use GDP per capita and labour productivity for the normality test (Shapiro-Wilk test)? Or should I also included GDP per capita squared and the interaction term?

[Nick Cox]

What do you intend to do that depends on marginal normality of any variable?

I've never seen a dataset with GDP per head that wasn't better analysed using its logarithm.

[Carlo Lazzaro]

Hester:
welcome to this forum.
As an aside to Nick's helpful comment, I would recommend you to follow the FAQ about posting (more) effectively (posting what you typed and what Stata gave you back is a very good first step to take. Thanks).
From your 1st post:
1) I'm not able to spot the regressand;
2) I do not understand if you mean -xtreg,fe-, -xtreg,re-, else;
3) I suspect that you have (inefficiently) created interactions by hand instead of relying in the wonderful capabilities of -fvvarlist-:

c.GDP##c.GDP

4) normality is a (weak) requirement for residuals distribution only.

[Hester vanderGaag]

Thank you Carlo.

For my thesis I use a panel data set. However, I am not sure which model I should use. My regression model is as follow:
Y = a0 + a1ln(x) + a2ln(x)^2 + a3 ln(z) + a4ln(x)*ln(z). The panel data set consist of 3 countries for 29 years.

My first question is: when I run a regression model for the panel data set as follows: xtreg Y ln(x) ln(x)^2 ln(z) ln(x)*ln(z), I each coefficients is significant. However, should I also do the hausman test to test whether I need a model with fixed effects or random effects or is my xtreg already sufficient since I have significant results?

Secondly: When should I also use robust or cluster(id)?

[Carlo Lazzaro]

Hester:
the first issue here is to switch to a different -xt command, as you have a T>N panel dataset.
See -xtgls- and -xtregar-.

[Hester vanderGaag]

Thank you Carlo!
When I'm using -xtgls- I get significant coefficients. However, I cannot distinguish between fixed effects and random effects, is that true? In addition, when I'm using -xtregar- I don't get any significant coefficient. Which one should I use? And should I use a Hausman test?

[Carlo Lazzaro]

Hester:
as per FAQ, please post what you typed and what Stata gave you back. Thanks.