Dear all,
I am attempting to estimate the impact of a program.
The program lasts for several years. Therefore, besides my baseline measure for my dependent variable (let's call it Y0), I have several follow-up measures of the same dependent variable (let's call them Y1, Y2, Y3, etc.)
As I work in a quasi-experimental setting, I use a data preprocessing method called entropy balancing. This method is analogous to other matching and weighting methods.
Once balanced (using the weights generated by entropy balancing), I estimate my treatment effect α with a simple linear regression, for example: Y3= α*treat + βi*predictori+ε
Treat is a dummy, taking the value one for individuals in the treated group, and 0 for individuals in the control group.
ε is the error term.
The baseline value Y0 is included in the entropy balancing weights AND the predictors.
The equation above estimates the effect of the program at time 3.
I wonder what it would mean to add the measure Y2 among the independent variables of the linear regression above. Intuitively, I feel like α would now express the effect of the program between time 2 and time 3.
That could very well be completely wrong. Besides, what if I add both Y2 and Y1 in the linear regression above?
Thank you so much for any insights!
Julie
I am attempting to estimate the impact of a program.
The program lasts for several years. Therefore, besides my baseline measure for my dependent variable (let's call it Y0), I have several follow-up measures of the same dependent variable (let's call them Y1, Y2, Y3, etc.)
As I work in a quasi-experimental setting, I use a data preprocessing method called entropy balancing. This method is analogous to other matching and weighting methods.
Once balanced (using the weights generated by entropy balancing), I estimate my treatment effect α with a simple linear regression, for example: Y3= α*treat + βi*predictori+ε
Treat is a dummy, taking the value one for individuals in the treated group, and 0 for individuals in the control group.
ε is the error term.
The baseline value Y0 is included in the entropy balancing weights AND the predictors.
The equation above estimates the effect of the program at time 3.
I wonder what it would mean to add the measure Y2 among the independent variables of the linear regression above. Intuitively, I feel like α would now express the effect of the program between time 2 and time 3.
That could very well be completely wrong. Besides, what if I add both Y2 and Y1 in the linear regression above?
Thank you so much for any insights!
Julie
Comment