Hi everyone,
I am working on a Fuzzy RD design and I have a question about the graphs I am doing.
Suppose I have the following data: w indicates treatment, z is the assignment variable, z0 is a dummy equal to 1 if z>=0, and y is an outcome variable.
I am estimating the effects doing something like this:
ivreg y (w=z0) z z2 z3 ...
I want to do a graph to show a discontinuity or not on the outcome variable.
I am doing this graph using the following code:
twoway
(lpolyci y z if z<0, degree(3))
(scatter y z if z<0)
(lpolyci y z if z>=0, degree(3))
(scatter y z if z>=0), xline (0)
Is this OK?
Or should it be:
twoway
(lpolyci y z if w==0, degree(3))
(scatter y z if w==0)
(lpolyci y z if w==1, degree(3))
(scatter y z if w==1, xline (0)
Any help would be greatly appreciated!
Alan
I am working on a Fuzzy RD design and I have a question about the graphs I am doing.
Suppose I have the following data: w indicates treatment, z is the assignment variable, z0 is a dummy equal to 1 if z>=0, and y is an outcome variable.
I am estimating the effects doing something like this:
ivreg y (w=z0) z z2 z3 ...
I want to do a graph to show a discontinuity or not on the outcome variable.
I am doing this graph using the following code:
twoway
(lpolyci y z if z<0, degree(3))
(scatter y z if z<0)
(lpolyci y z if z>=0, degree(3))
(scatter y z if z>=0), xline (0)
Is this OK?
Or should it be:
twoway
(lpolyci y z if w==0, degree(3))
(scatter y z if w==0)
(lpolyci y z if w==1, degree(3))
(scatter y z if w==1, xline (0)
Any help would be greatly appreciated!
Alan