Within estimator estimate

[Raja Hasan]

I have the following equation:

in my data I have year and Id

I want to get the within the parameter estimates of the above regression
Which one of the following would be correct and why?

xtreg A L1.A L1.B L1.C yr*, fe or

xtreg A L1.A L1.B L1.C, fe or

xtreg A B C yr*, fe or

xtreg A B C, fe


Can anyone help me? Should I include vce(robust)? If not, why?

[Carlo Lazzaro]

Raja:
if you have a lagged dependent variabe as a predictor, I think you should go dynamic panel data regression (see -xtabond-).

[Raja Hasan]

Hi Carlo,

I will do that but first I have to find within parameter estimates. Could you please tell me how I can do that?

[Carlo Lazzaro]

Raja:
I fail to get how could I help you out in thus respect, as I do not know your data.

[Phil Bromiley]

To expand on Carlos's comments, I have no idea what the OR option means in xtreg. Whether you should use the current values of the right-hand side variables or lagged values of the right hand side variables is a matter of substance on which we cannot comment since we have no understanding of your context.

Fixed effects give you the within estimates.

As Carlos pointed out, with the lagged dependent variable, depending on your data, you may need to take care of the potential endogeneity of that lagged dependent variable with either one of the xtabond or related estimators or xtdpdml (user written).

Whether robust standard errors are necessary depends on whether you think the error variance will vary across panels or across time within panels. Many use robust standard errors as a matter of standard practice.