Hello,
I am running a difference-in-differences model. The time (year) of the treatment is unique for all the observations. The treatment variable is continuous. I use the command below to obtain the yearly treatment effects. Excluding the reference year, there are 3 pre-periods and 5 post-periods.
reghdfe y b4.year##c.x, absorb(i t) cluster(i t) noomit nocons
I would be grateful if you could help me with the following:
1. The coefficients before the treatment are not statistically significant. However, I would like to further check if the parallel trend assumption is likely to hold by using the pretrends command (Roth, 2022) to evaluate the power of conventional pretests and the distortions from pre-testing in the data-generating process. However, I did not understand how to use this command. Could you explain to me how to proceed? Also, do you know other tests and commands that I can use?
2. Since my treatment is continuous, Callaway et al. (2021) show that the validity of my approach requires what they call the “strong parallel trend” assumption. Do you know any command that allows me to test whether the "strong parallel trend" assumption holds? Also, it would be useful if you know any command that allows me to run a difference-in-differences model using the estimator that they propose or if you know any other estimator that is valid in my setting.
Thank you very much in advance!
I am running a difference-in-differences model. The time (year) of the treatment is unique for all the observations. The treatment variable is continuous. I use the command below to obtain the yearly treatment effects. Excluding the reference year, there are 3 pre-periods and 5 post-periods.
reghdfe y b4.year##c.x, absorb(i t) cluster(i t) noomit nocons
I would be grateful if you could help me with the following:
1. The coefficients before the treatment are not statistically significant. However, I would like to further check if the parallel trend assumption is likely to hold by using the pretrends command (Roth, 2022) to evaluate the power of conventional pretests and the distortions from pre-testing in the data-generating process. However, I did not understand how to use this command. Could you explain to me how to proceed? Also, do you know other tests and commands that I can use?
2. Since my treatment is continuous, Callaway et al. (2021) show that the validity of my approach requires what they call the “strong parallel trend” assumption. Do you know any command that allows me to test whether the "strong parallel trend" assumption holds? Also, it would be useful if you know any command that allows me to run a difference-in-differences model using the estimator that they propose or if you know any other estimator that is valid in my setting.
Thank you very much in advance!
Comment