Does anyone know how to compute a test of over-identifying in a system of simultaneous equations? Sargan and Hansen tests are just used for a single equation but I need the test for the whole system of many simultaneous equations. I think there is a Hansen-Sargan test for this but I did not find any explicit reference exposing the formula.
Hello Ermira, I am facing the same problem and have found no answer
check
https://statistics.stanford.edu/sites/default/files/AND%20NSF%2036.pdf
An equation is identified if the number of absent exogenous variables of the equation is greater than or equal to the number of endogenous variables present in the equation, minus one.
Over identifying may be opposed in your program by an averaging of more than one variable. This means that elimination a variable will not tell you much about the over-identification situation. The most obvious way of checking is to count the number of identifying equations and the number of variables. If and when the latter exceed the former, which of the variables to eliminate is then a matter of your sense and choice.
I am so grateful with your prompt responses....thanks to all of you!
An overidentified simultaneous equation model can be perfectly and easily estimated by 2SLS method. It is a very simple but a significant method by which one can test the significance of partial coefficient and overall significance of a model too.
- Badges
- Science topic
- Similar topics
- Political Science