[Carlo Lazzaro]

Siege:
I think you have to live with the unbalanced panel dataset (a frequent fact of life) and go on with your regression.

[Siege Taker]

H there,

Please do you know of tests/commands I can use to assess which independent variables I have to drop?

Cheers

[Carlo Lazzaro]

Siege:
your goal should be to give a fair and true view of the data generating process, no matteri if the predictors end up significant or not.

[Siege Taker]

Carlo,
I have achieved that goal now, do you have any idea on how I can cut down on the 52 independent variables? There are diverse opinion in the literature as to their significance

[Carlo Lazzaro]

Siege:
I assume that the literature in your research field reports a core set of predictors that have to be included in the regression model in order to give a fair and true view of the data generating process.
Start from that set and add controls (if any), keeping in mind that the more predictors you include, the harder your goal to explain/disseminate the results of your research.

[Siege Taker]

Hi there,

Please do you know if GMM is suitable for panel data set with a categorical dependent variable? a dummy variable has been assigned to the categories 0,1 and 2

[Siege Taker]

Carlo Lazzaro
You once recommended xtologit with clustered standard errors to me. Please do you know if the proportional odd assumption applies to xtologit as it is the case for ologit?
Secondly, can an endogenous variable be controlled using any options in xtologit?

Thank you

[Carlo Lazzaro]

Siege:
I think that the proportional odd assumption makes sense for -xtologit-, too.
I'm not aware of any command to deal with endogeneity in panel ordered logistisc regerssion.