TSLS vs GMM

[Apple Tan]

I need to run an IV

1) How to decide whether to use TSLS or GMM?

2) For GMM, I note that it doesn't have F-stat. So how do I tell whether the instrument is strong?

3) For both TSLS and GMM, how do I tell the instrument is exogenous?

Thank you

[startz]

TSLS is a special case of GMM

[startz]

And you generally need to use economic theory to tell if an instrument is exogenous

[Apple Tan]

Hi

Thank you for the replies.

I still don't understand.
1) When should I use GMM vs TSLS? Anywhere that I can read up more about this

2) For GMM, there is no F-stat. So how do I tell if the instrument is weak? Usually, a F-stat<10 implies weak instrument

Thank you

[Apple Tan]

Hi

1) I still don't understand how to decide between TSLS & GMM.

Ans:
2) Ok, I found out for GMM (non-panel data), the F-stat can be found at View --> IV Diagnostics and Tests

Query:
1) GMM / DPD - Generalized Method of Moments / Dynamic Panel Data:

i) How do I find the p-value for the J-stat?
ii) How do I find the F-stat?

2) For a panel data, using TSLS

i) how do I find the J-stat and its probability?

Thank you

[startz]

If the model is linear in the parameters and you think the errors are iid, then use 2sls. Otherwise GMM might be better.

You should not necessarilyrely on F>10 as a test for weak instruments if you have more than 1 endogenous variable.