Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Optimal RD Bandwidth Choice also for Rectangular Kernel?

 From andreas nordset To statalist@hsphsun2.harvard.edu Subject Re: st: Optimal RD Bandwidth Choice also for Rectangular Kernel? Date Thu, 26 May 2011 18:04:57 +0200

```Great, thank you so much!
AN

On Thu, May 26, 2011 at 5:30 PM, Austin Nichols <austinnichols@gmail.com> wrote:
> andreas nordset <andreas.nordset@gmail.com>:
>
> You can multiply by 5.4/3.4375 or 1.57,
> or you can specify the uniform option and have their program do it for you.
>
> On Thu, May 26, 2011 at 10:54 AM, andreas nordset
> <andreas.nordset@gmail.com> wrote:
>> Thanks so much Austin!
>>
>> It says h_opt = CK times other things, so once I have the h_opt for
>> the triangular kernel, I will just divide by the CK=3.4375 that they
>> say is the CK value for the triangular edge kernel, and multiply
>> instead by the CK applicable to the uniform kernel,
>> where I should be able to compute the latter from the formulas given then?
>>
>> Thanks again and all best,
>> AN
>>
>>
>> Austin Nichols <austinnichols@gmail.com>:
>>
>>> andreas nordset <andreas.nordset@gmail.com>:
>>>
>>> The two differ by a constant as explained in the WP; in the code look for
>>>
>>> if "`uniform'"~="" {
>>> local CK = 5.4
>>> }
>>> else {
>>> local CK = 3.4375
>>> }
>>>
>>>
>>> On Wed, May 25, 2011 at 12:37 PM, andreas nordset
>>> <andreas.nordset@gmail.com> wrote:
>>>> Dear Statalist members,
>>>>
>>>> I would like to use the -rdob- Optimal-Bandwidth RD estimator by
>>>> Fuji-Imbens-Kalyanaraman,
>>>> as posted at http://www.economics.harvard.edu/faculty/imbens/software_imbens
>>>> together with the relevant documentation / help file,
>>>> and based on [Imbens, Guido and Karthik Kalyanaraman, "Optimal
>>>> Bandwidth Choice for the Regression Discontinuity Estimator", NBER WP
>>>> 14726, Feb. 2009].
>>>>
>>>> Now, if I understand the documentation there correctly, the algorithm
>>>> it uses is based on the use of a triangular kernel, and I understand
>>>> the documentation as saying that which bandwidth is optimal will vary
>>>> with the kernel chosen.
>>>> At the same time though, I understand Imbens-Lemieux 2009 [Imbens,
>>>> Guido and Thomas Lemieux, "Regression discontinuity designs: A guide
>>>> to practice", Journal of Econometrics 2008, as suggesting that a
>>>> sensible baseline specification is to use a simple rectangular kernel
>>>> and then just check sensitivity to the bandwidth.
>>>>
>>>> So I'm wondering whether any of you could think of a way to find out
>>>> what the optimal bandwidth would be when using a rectangular rather
>>>> than a triangular kernel, short of writing an alternative to the
>>>>
>>>> Thank you so much and best regards,
>>>> AN
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```