[Feature Request]: RDD with user-defined bandwidth

[chiara-fb]

Describe the feature you want to propose or implement

Hi,

The class RDFlex does not allow for user-defined bandwidth. If initialized with a fixed h value, it throw the following error:

rdrobust.rdrobust.rdrobust() got multiple values for keyword argument 'h'

Would it be possible to extend the class to allow for fixed h?

Propose a possible solution or implementation

No response

Did you consider alternatives to the proposed solution. If yes, please describe

No response

Comments, context or references

No response

[OliverSchacht]

Hi @chiara-fb !

Thanks for pointing out this issues.

It is currently only possible to set h_fs and then fit(n_iterations=1) in order to control the bandwidth of the ML adjustment.
The final estimation bandwidth is then still set automatically and cannot be controlled.

We are planning on adding the requested feature and allow the bandwidths h and b to be fixed.
However, using these kwargs could have negative consequences, especially when the first stage bandwidth is smaller than h. We will add a warning for this.

We will add this in the next days and notify you here!

[OliverSchacht]

Hi @chiara-fb - the requested feature is available in #343

[chiara-fb]

Thank you Oliver!