EViews.com

Good afternoon, everyone!

I've run 4 similar GMM models, in which the only difference was the response variable. For instance, in the first one, the response variable was the Real Exchange Rate. On another one, it was the Real Effective Exchange Rate.
Anyway, regardless of what the response variables are, I was wondering how I can choose between the models to be the most robust. What I mean is, is there any criteria of choice to pick out the best one of them 4? I was choosing by looking at the Sum Squared Resid since we cannot look to the R² on GMM modeling.

Can anybody help me with this?

Thanks!

Could anyone please help me?

Good afternoon, everyone!

I've run 4 similar GMM models, in which the only difference was the response variable. For instance, in the first one, the response variable was the Real Exchange Rate. On another one, it was the Real Effective Exchange Rate.
Anyway, regardless of what the response variables are, I was wondering how I can choose between the models to be the most robust. What I mean is, is there any criteria of choice to pick out the best one of them 4? I was choosing by looking at the Sum Squared Resid since we cannot look to the R² on GMM modeling.

Can anybody help me with this?

Thanks!

Could you explain what it means to be the most robust? If you have one model that predicts apples and another that predicts pears, what does it mean to say one model is better than the other?

Good afternoon, everyone!

I've run 4 similar GMM models, in which the only difference was the response variable. For instance, in the first one, the response variable was the Real Exchange Rate. On another one, it was the Real Effective Exchange Rate.
Anyway, regardless of what the response variables are, I was wondering how I can choose between the models to be the most robust. What I mean is, is there any criteria of choice to pick out the best one of them 4? I was choosing by looking at the Sum Squared Resid since we cannot look to the R² on GMM modeling.

Can anybody help me with this?

Thanks!

Could you explain what it means to be the most robust? If you have one model that predicts apples and another that predicts pears, what does it mean to say one model is better than the other?

Startz, thanks for answering!
Anyway, the four response variables I used are supposed to tell us a similar information, since they're all exchange rates. But what I was really trying to understand is if there's a way to say which model explains better the response variable. In other words, given that the independent variables are the same, which response variable is being more well explained by those variables. Do you get it now?

If you have different models that all explain apples, then there are some standard methods of comparison. You pointed out R^2, for example. If the explained variables are different, there is no standard method.

Got it, Startz. Thanks a lot!

If you have different models that all explain apples, then there are some standard methods of comparison. You pointed out R^2, for example. If the explained variables are different, there is no standard method.

Now what if I wanted to use 4 different explanatory variables to explain 1 response variable at a time? That is, I have the same 4 models, but now the independent variables are not the same. What would be the best criteria? I know that R^2 is not the best one when it comes to GMM.

There's still not a magic rule, but you might well want to choose according to the AIC or the BIC.