Hello,
I writing a thesis on the relationship between income inequality and economic growth, in USA. My question is how many observations do I need for a regression analysis. I was planning on finding data per state, for 20-30 years but it was impossible. Thus, now I switched to annual USA data (as quarterly was also difficult to find) and I extended my time span to have the maximum observations possible. I ended up with 60 observations per variable. Is it okay?
dear Kostas Simoglou
I think 60 observations are enough for a regression analysis to get accurate results if you apply Autoregressive Distributed Lag (ARDL) cointegration
best regards
It depends on the technique, etc. you use.
Furthermore, there is no clear number. What happens is that some techniques requires assuming that the number of observations (or the number of unit in a certain dimension) tends to "infinity" (which could be 30-50).
For instance if you have observations per state and you have 50 states and, in addition to that, you have several observations per stata, this should be enough for almost every analysis.
It also depends on how many variables you have for estimation. you will lose more degree of freedom if you have many independent variables in your model.
Thank you all for your answers, much appreciated. @Sayyed Mahdi Ziaei I have 4-6 independent variables. Is it okay?
It is okay. First, run your model and then, check it for diagnostic tests. If you have more observations, the chance of getting the better result is higher. Good luck!
Numerous researchers say that there should be at least 10 observations per variable. If we are using five independent variables, then a clear rule would be to have a minimum sample size of 50.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis