Although I am not actually carrying out panel data after PSM, I'm exploring its possibility. I do have matched data after propensity score matching. But my point is: if I obtain matched data for each year and then merge the data (such that the data contains both person identification as well as time i.e. years), how can I carry panel data analysis? Should I control only for propensity scores or should I include other variables as in the case of regression adjustment? My doubt is: since the variables are already matched, would it still be possible to carry out panel data analysis including all the required variables or not?
Try to read this Econometrics book; An Introduction to Modern Econometrics Using Stata, by Chris Baum. Its very rich on panel data analysis.
www.ncer.edu.au/events/documents/Singapore_lecture1.pdf
www.amazon.com/Introduction-Modern-Econometrics.../1597180130
The short answer: yes you can. Just make sure that you match individuals (all years) rather than yearly observations for each individual. Also, it is best to only match on pre-treatment characteristics, if possible.
@Bonga-Thank you for the documents. @Jonathan- Thank you for the response. So, you mean, I append data for all years followed by matching and then conduct panel data analysis?
@Jonathan- One more thing I want to be clear is about the use of X variables in panel data analysis. Since I have propensity scores obtained after PSM, what do you think is appropriate to adjust in panel analysis- just the propensity scores or the set of X variables?
You can use any variables in the panel model after PSM. When you do, it adjusts for any remaining differences. If the matching removed all differences, then including any variables other than the treatment indicator should not change the estimate of treatment effects.
Thank you Jonathan. Will be equally grateful to get some references regarding this.
The term used with studies that implement PSM followed by regression is "double-robust". Here are a few good references:
Kang, Joseph DY, and Joseph L. Schafer. "Demystifying double robustness: A comparison of alternative strategies for estimating a population mean from incomplete data." Statistical science (2007): 523-539.
Emsley, Richard, et al. "Implementing double-robust estimators of causal effects." Stata Journal 8.3 (2008): 334-353.
Funk, Michele Jonsson, et al. "Doubly robust estimation of causal effects." American journal of epidemiology 173.7 (2011): 761-767.
Sasidharan, Lekshmi, and Eric T. Donnell. "Application of propensity scores and potential outcomes to estimate effectiveness of traffic safety countermeasures: Exploratory analysis using intersection lighting data." Accident Analysis & Prevention 50 (2013): 539-553.
I hope this helps!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression