Hello,
According to my research, missing data can be imputed based on the type of missingness, and when analyzing the data, choosing just one of the imputed datasets is not appropriate—you should use all of them. However, when I use imputed datasets in my tests or analyses, 9 times out of 10 I get an error saying that my data is mi (multiply imputed), and that a certain command cannot be used with imputed data.
In addition, the countries and years have been imputed, so I’m always stuck. Even if I find alternative commands or methods, an error always pops up because the data is imputed.
Is there anything I can do about this? I just want to identify the long- and short-term relationships between my variables, with N = 16 and T = 17!
In addition, I have cross-sectional dependence (CSD) in my tests, so I need to use second-generation unit root and cointegration tests, like Westerlund. However, I can’t even run them in Stata because two of my variables have 11 and 10 missing values, respectively.
I’m considering using CS-ARDL and then applying CCE-AMG to check the robustness of the results—although I’m not entirely sure how that would work.
If my main goal is to examine the relationship between my variables, is it necessary to test for heteroskedasticity and autocorrelation? If so, how do I correct for that using the methods I’m planning to use?
But first... how do I fill in my data and make xtdcce2 work?!
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA