mixed command for fixed effect model

[Roxanna Sh]

Hello,

I developed a random effect model as follows:

mixed firmperformance_ln typeofstrategy firmage firmsize i.y* || id: ||country:

I am interested in understanding the role of strategy in firm performance and other variables are control variables.

Now I want to write fixed-effect model I used the menu of the Sata and also I read the help of the command and my model is


mixed firmperformance_ln typeofstrategy firmage firmsize i.y*


But I am not sure that is correct or not. The nested part is just omitted in this model

Is this model correct?

Thank you

[Carlo Lazzaro]

Roxanna:
1) youc coded a three-level random intercept -mixed- model. Its fixed portion (i.e., the one that you reported as your second code) ends at the left of the first double pipe sign;
2) if you want to go -fe- with (I surmise) a panel dataset with a contnuous regressand, you may want to try something along the following lines (that, however, ignore the nested design of your study):

xtset firm year
xtreg  firmperformance_ln typeofstrategy firmage firmsize i.y*, fe (vce(cluster firm)

As usual, please note that panel needs to be at least 30 (rule of thumb) to make clustered standard errors work as expected.

[Roxanna Sh]

Thanks. My observations are about 60,000. Is it possible to use mixed command for fixed-effect model instead of xtreg? And if yes, I do not know how I should write the model?

[Andrew Musau]

In a linear model, you can always include fixed effects using indicators, but this is a very inefficient way of estimating the model. The mixed syntax is

mixed depvar fe_equation [|| re_equation] [|| re_equation ...] [, options]

If individuals are nested in countries, then country fixed effects are collinear with individual fixed effects and you just need to include the latter.

mixed firmperformance_ln typeofstrategy firmage firmsize i.y* i.id