What is the minimum number of observations with which a GMM Panel Data estimation technique can be carried out? My total observations is 104
Please take care with number of variables to be included and lags as well!
You have quite a sufficient and statistically significant number of observations.
GMM is characterized by small t and large n, so n in most cases should be greater than t. An important issue with GMM is the problem of 'Too Many Instruments', since the number of instruments should less than the number of the cross-sections, which was highlighted by Roodman (2009).
Its an unbalanced panel with my time dimension ranging from 3-5 years. For some countries I have data for 3 years, for some 4 years and for most of them there is data for 5 years. This is hinged on the availability of data for my Dependent Variable ( Fiscal Transparency) .May I also point out that survey from which data for my dependent variable is derived is taken only once in two years. i.e my time dimension is for 2006, 2008, 2010, 2012 and 2015. I don't know if this non-consecutive nature of the time dimension is also contributing to my inability to run the estimations.
I suggest to you to use the collapse function in STATA to reduce the number of instruments. You use the "collapse" function within gmm or gmm style endogenous system equation.
GMM can’t be applied in less numbers of observation. Atleast 200 observations are required in your case.
- Badges
- Science topic
- Similar topics
- Mathematics