Hello,
I have several questions regarding the fundamental approach of my regression model.
I have a panel dataset, with n=173 countries and t=12 (2004-2015). The panel is strongly unbalanced, since it doesn't include observations for all countries at each year. In general, I try to measure the effect of public finance investment in solar energy (main independent variable) on private finance investment in solar technology (Dependent variable). Besides, I use several different energy and macro-related variables. Some are logarithmized (here indicated as _ihs_ as I use inverse hyperbolic sine transformation to retain all zero observations). But none of my explanatory variables are dummies .
I use an existing paper on the similar topic as a benchmark. They use a nearly identical dataset and panel structure as well, while facing the same issue of an unbalanced dataset (see Screenshots 1&2 from the original paper). In their paper, they do not use the intuitive approach of testing with Fixed Effects (or random effects) but use the "pooled OLS/WLS and FE regression model with Driscoll Kraay standard errors". This approach goes by the command - xtscc depvar indepvar, lag(..) - . Screenshot 3 is an example of my personal regressional models.
My main question 1 is, can I use this Driscoll-Kraay estimator over FE to conduct a regressional model with meaningful, consistent and efficient results?
Question 2: if Q1 is answered with yes, am I using this command correctly - and is the factor "time" (<- here years) included?
Question 3: if comparing my results from Driscoll-Kraay estimation to the regular OLS (photo 4) or FE (photo 5) model, the statistical significanes are much higher and rather apply to my general assumptions over the influence of all x-variables on y. How can I explain these differences?
Please comment, if my questions are hard to comprehend and require more input. I am happy about every help, since I target to come up with a rational argument, why I am using this Driscoll-Kraay estimator. Therefore, I need more fundamental understanding to differentiate my approach from e.g. Fixed Effects.
Every help is greatly apreciated, since I am not very deep into econometrics, but give my best.
Array
Array
Array
Array
Array
Related Posts with Fundamental Question on Panel regression Pooled OLS (Driscoll-Kraay Est.) and Fixed Effect
Panelregression using pooled ols with dummiesHello all, I have a paneldataset ranging from 2002 until 2019 containing ~ 500 companies. My paneli…
multiple imputation (-mi-) with the synthetic control method (-synth-)Has anyone combined multiple imputation with the synthetic control method (https://fmwww.bc.edu/repe…
Replacing missing values in a panel dataHello, I am doing a difference in difference in panel data set, years 2001-2015. The observation un…
P-values and CI's for dotplot?Hi you helpful people. I have measured a concentration of a medical drug 1 week after operation in …
Rearranging Data Code: clear input str25 sub1 marks1obt marks1max str30 sub2 marks2obt marks2max str30 sub3 marks3ob…
Subscribe to:
Post Comments (Atom)
0 Response to Fundamental Question on Panel regression Pooled OLS (Driscoll-Kraay Est.) and Fixed Effect
Post a Comment