rdrobust command bandwidth usage

[Muza Gadimova]

Hello,

I'm using rdrobust command to conduct an RDD design for my master's thesis but I am having problems with its implementation. It seems like it does the regression on the whole dataset no matter which bandwidth type I specify in options h(). I attached my command and the results to this post. As you can see, the number of observations within the bandwidth is too big. This is actually the number of observations without any bandwidth. I hope I could explain what the problem is and would really appreciate if could get some help. Thanks in advance.

[Jared Greathouse]

What are mass points in the running variable and why are they a problem?

[Muza Gadimova]

The running variable is date of birth, so I figured it is normal to have mass points in the running variable. Should that be considered as a problem?

[Øyvind Snilsberg]

"Eff. Number of obs" shows how many observations you are using for a given bandwidth. Typing e.g. -count if rv1 > 0 & rv1 <= 64.510- should return 1159, the number of observations above the cutoff within the bandwidth.

[Muza Gadimova]

"Eff. Number of obs" shows how many observations you are using for a given bandwidth. Typing e.g. -count if rv1 > 0 & rv1 <= 64.510- should return 1159, the number of observations above the cutoff within the bandwidth.

Does that mean that the results of the regression are already based on the observations in a given bandwidth, not the whole dataset of 8395?

[Øyvind Snilsberg]

Does that mean that the results of the regression are already based on the observations in a given bandwidth, not the whole dataset of 8395?

That is correct.