Panel Data

[Arpita Dey]

hello everyone,
Currently, I am working with panel data set with N=7 and T=10. I ran the regression using xtreg and the random effect model is giving me a better result. However, I wanted to check heteroscedasticity. hence I simply regress and did a hettest and it was showing that the null hypothesis is rejected. Now I want to know how to remove it using panel data

[Carlo Lazzaro]

Arpita:
welcome to this forum.
You may want to take a look at the following threads:
https://www.statalist.org/forums/for...random-effects
https://economics.stackexchange.com/...vs-time-series

In addition, with N=7 and T=10, you are reaching the switch point between -xtreg,re- and -xtregar,re- or -xtgls-, when appropriate.

[Arpita Dey]

Dear carlo,
I have tried xtreg dep var indpen var,re roubust.where One of the independent variable showed positive relation.
Also ,I have xtgls where the above-mentioned variable showed a negative relation .which one should I rely on?

[Carlo Lazzaro]

Arpita:
I'd go -xtgls- with the standard errors corrected for heteroskedasticity, if you detected it.

[Arpita Dey]

Thanks for your advice.
I want to know If I use xtgls , igls panels(heteroskedastic) ,then for interpretation which model will fit in for xtgls?

[Carlo Lazzaro]

Arpita:
T>N panel datasets via feasible generalized least squares.

[Arpita Dey]

↵

• I am afraid.One of the variable is showing unrealistic relationship under xtgls.And another thing is showing name Iteration tolerance .why is it happening ?

[Carlo Lazzaro]

Arpita;
as per FAQ, please post what you typed and what Stata gave you back with -xtreg,re- and -xtgls-. Thanks.

[Arpita Dey]

This is what I typed and stata gave me back this .In the second image,my inflation variable is showing negative relationship with respect to crime rate. But when I ran normal xtreg,re where it showed positive although insignificant in both model.

[Carlo Lazzaro]

Arpita:
due to its wide coefficient interval, the coefficient of your inflation variable is negative but itr real, fixed and unknown value can well be 0 or something else in the population from which your sample was drawn.

[Arpita Dey]

Thanks for your answers. So should I stick with this result? As it is a matter of my thesis final findings. I am a little bit puzzled

[Carlo Lazzaro]

Arpita:
results are what they are.
If your model is correctly specified, you should stick with its results.

[Nick Cox]

Although there are many nuances and subtleties here, the big picture seems to be that you're fitting a complicated model to just 70 data points. Something has to give. Either you simplify the model and are likely to be criticised by your teachers for not trying out possibly important predictors. or you appreciate that such a fit is quite a strain.

[Arpita Dey]

On the other hand when I am using only xtgls,panels(hetero) instead of xtgls,igls panels(hetero) ,inflation is showing positive relationship on crime.Why is this happening ? which one is more appropiate.

[Carlo Lazzaro]

Arpita:
does the 95% CI of the .inflation- coefficient include zero?