Estimating Fixed Effects in Panel Data with Time Lag

[Girlan Oliveira]

Hi, Statalists.

I'm working with panel data composed of 27 states and 12, but have a lag of year as the period of my analysis is 2001-2013 excluded the year 2010, because this year I have no data.
I'm estimating models of fixed effects and before that checked through the commands "xttest3", "xtserial" and "xtcsd, Pesaran abs" the existence of heteroscedasticity, serial correlation and contemporaneous correlation, respectively, in my data.
I'm having trouble to correct these problems before proceeding with the estimation of fixed effects, because:

1 - I do not know how to use Stata in the HC4 estimator to solve the problem of heteroscedasticity;

2 - Because of year lag can not use the "xtscc" to solve the problem of contemporary correlation;

3 - I do not know how to solve the second problem of serial correlation using the stata.

I hope you can help me.

Thanks in advance

[Carlo Lazzaro]

Girlan:
asuming that you have a continuous depvar, with a large N, small T panel data like yours you can go -xtreg, fe- with -vce(robust) (which, for panel only, works as the same as -vce(cluster clusterid)- )for dealing with both heteroskedastocity and serial correlation.

[Girlan Oliveira]

Thank you very much, Carlo.

Can you tell me something about how to solve the problem of contemporaneous correlation, which is also called cross-sectional dependence. As I said, I found some papers that indicate the use of the command "xtscc" but Stata returned me the following error message, due to the time gap that I had already mentioned:

ano is not regularly spaced: there are contemporaneous gap (s) across all subjects in uf

OBS.: "ano" is my time variable and "uf" is the variable that represents the states.

[Carlo Lazzaro]

Girlan:
As far as I can get from the -help xtscc- file, it applies to small N, large T panel dataset (seemingly the opposite of what you're dealing with).