Hi
I am studying the panel causality between two variables (EX. growth and trade) for large sample of countries. I subdivide the countries based on their regime (many criteria for regime) … let say I want to test the causal relation between growth and trade on two regimes (Ex. high financial openness countries and low financial openness countries). However, some countries have the two regimes during period of Study (Ex. 22 years).
If I split the sample to two groups based on regime (high and low financial Openness). The cross sections on each group will not have the same no. of T.
1. Is this rise any problem in estimation?
2. If the T is not the same for cross sections on my sample, I cannot go for cross-section dependence test. And later for 2nd generation panel data tests. How I can resolve such problem?
Thank you.
You have something like two panel that are homogenous and than want to compare growth-trade nexus between groups?
If you have something like that on your mind that spliting means you will estimate two panel and you will make your conclusion based on the differences.
Robustnes of each estimation could depend on many factors but first what is the number of your cross section units?
for other questions I'm not sure to answer and I'm afraid my answers cause misunderstanding, but you can test cross-sectional dependency using Pesaran's test in Stata software and using xtcsd command after xtreg. xtcsd with option Pesaran can handle balanced as well as unbalanced panels.
bests...
Ehsan
Thank you Bojan and Ehsan
I have Four groups with Ns: 58, 48, 47, 26
Each group will be splitted for two samples (HFO, and LFO).
For example if we take group with 58 cross-sections, and divide it to 2 samples....> some cross sections will be in both samples, but the year of course different. as some countries has HFO all the period, and others have HFO, and LFO in thier period.
Therefore, If we splitte the group, some cross section will have 22t , others may 15, 10, 8, ...etc. not all the same. depending if they was HFO on that year or not.
Hope this will be clear for the problem
thanks for all your feedbacks
If you set your analysis in such way that you compare two or more different groups that are homogenous that is nice way out of lot of complication that can appear in panel. I would not worry if your groups have differen N or T, but I would worry if my T and N large enough
There is Panel Handbook, Oxford press is a publisher. You should look at it.
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India