Dealing with missing values in a regression analysis

[John Cavin]

Hello,

I am using Stata12 and would like to regress by using lagged measures and include two fixed effects (say time and style). I have tried following:

xi: reg y laggedy x2 x3 x4 x5 i.year i.style

Questions:
1) How do I have to deal with missing values (e.g. "laggedy" has more than 1000 missing values)? Is my formula right?
2) I would like to add fixed effects: i.year & i.style. Do I have to add anything?
3) How can I cluster standard deviation?

If necessary, I can give you further information.
Thanks a lot!

[Marcos Almeida]

Hello John,

To start, and trying to answer 3 of your queries:

Stata deals with linear regression (under - regress - ) by using listwise deletion.

You don't need to apply - xi:- to work with the dummies in recent versions of Stata, neither create a lagged "Y" variable (for this you just need to apply the lag operator, that is. "L.").

Hopefully that helps.

Best,

Marcos

[John Cavin]

Hello Marcos,

thanks a lot for your answer! So I do not need to add any code for missing values?
And what about fixed effects? Is it right, that following includes fixed time and fixed style effect?

reg y L.y x2 x3 x4 x5 i.year i.style

[daniel klein]

As has been repeatedly discussed on the list and is well known from the literature, the estimates obtained from this model will be biased by design. See

help xtabond

for a proper estimator of models with lagged responses (dependent variables).

Best
Daniel