Rdrobust help with bandwidths

[Lindsey Buck]

Hi all,
I am trying to get rdrobust results with the mserd bandwidth, half of it, and double it. I am running the following code. It tells me that bwselect is incorrectly specified. Any ideas on this? Thanks!

rdbwselect call_accept riskx, kernel(tri) p(1) c(0) bwselect(mserd)
local bandwidth=e(h_mserd)
local bandhalf=`bandwidth'/2
local banddouble=`bandwidth'*2
rdrobust call_accept riskx, p(1) c(0) bwselect(mserd)
outreg2 using "table2.xls", replace
rdrobust call_accept riskx, p(1) c(0) bwselect(`bandhalf')
outreg2 using "table2.xls", append
rdrobust call_accept riskx, p(2) c(0) bwselect(mserd)
outreg2 using "table2.xls", append
rdrobust call_accept riskx, p(2) c(0) bwselect(`banddouble')
outreg2 using "table2.xls", append

[Sebastian Calonico]

In rdrobust, to manually select the bandwidth you should use the h() option:

rdrobust call_accept riskx, p(1) c(0) h(`bandhalf')
rdrobust call_accept riskx, p(2) c(0) h(`banddouble')

[Cat Santos]

Hello,

I have a followup question that I kindly would like to ask.

I am interested in the robust estimates given by rdrobust, as these are the ones typically required by the journals I submit to. However, I was wondering if in the robustness checks to manually contract the bandwidth, I should change b() instead of h().

My question arises because if I run my normal model, I get:

. rdrobust dv1 running, c(0) all kernel(tri) weights(w)
Multicollinearity issue detected in covs. Redundant covariates were removed.
Mass points detected in the running variable.

Covariate-adjusted Sharp RD estimates using local polynomial regression.

   Cutoff c = 1944 | Left of c  Right of c            Number of obs =       1263
-------------------+----------------------            BW type       =      mserd
     Number of obs |        64        1199            Kernel        = Triangular
Eff. Number of obs |        34          58            VCE method    =         NN
    Order est. (p) |         1           1
    Order bias (q) |         2           2
       BW est. (h) |     4.385       4.385
       BW bias (b) |     5.930       5.930
         rho (h/b) |     0.739       0.739
        Unique obs |        11          54

Outcome: dv1. Running variable: running.
--------------------------------------------------------------------------------
            Method |   Coef.    Std. Err.    z     P>|z|    [95% Conf. Interval]
-------------------+------------------------------------------------------------
      Conventional | -.13807     .05677   -2.4318  0.015   -.249342     -.026789
    Bias-corrected | -.30086     .05677   -5.2992  0.000   -.412138     -.189584
            Robust | -.30086      .1136   -2.6484  0.008   -.523518     -.078203
--------------------------------------------------------------------------------

However, if I insert the optimal bandwidth given by this model into the h() option, I do not get the same results:

. rdrobust dv1 running, c(0) all kernel(tri) weights(w) h(4.385)
Multicollinearity issue detected in covs. Redundant covariates were removed.

Covariate-adjusted Sharp RD estimates using local polynomial regression.

   Cutoff c = 1944 | Left of c  Right of c            Number of obs =       1263
-------------------+----------------------            BW type       =     Manual
     Number of obs |        64        1199            Kernel        = Triangular
Eff. Number of obs |        34          58            VCE method    =         NN
    Order est. (p) |         1           1
    Order bias (q) |         2           2
       BW est. (h) |     4.385       4.385
       BW bias (b) |     4.385       4.385
         rho (h/b) |     1.000       1.000

Outcome: dv1. Running variable: running.
--------------------------------------------------------------------------------
            Method |   Coef.    Std. Err.    z     P>|z|    [95% Conf. Interval]
-------------------+------------------------------------------------------------
      Conventional | -.13807     .05716   -2.4156  0.016   -.250088     -.026044
    Bias-corrected |  -.0576     .05716   -1.0077  0.314    -.16962      .054425
            Robust |  -.0576     .07757   -0.7425  0.458   -.209637      .094442
--------------------------------------------------------------------------------

It's only when I use b(4.385) that I do get the same results:

rdrobust dv1 running, c(0) all kernel(tri) weights(w) b(4.385)
Multicollinearity issue detected in covs. Redundant covariates were removed.
Mass points detected in the running variable.

Covariate-adjusted Sharp RD estimates using local polynomial regression.

   Cutoff c = 1944 | Left of c  Right of c            Number of obs =       1263
-------------------+----------------------            BW type       =      mserd
     Number of obs |        64        1199            Kernel        = Triangular
Eff. Number of obs |        34          58            VCE method    =         NN
    Order est. (p) |         1           1
    Order bias (q) |         2           2
       BW est. (h) |     4.385       4.385
       BW bias (b) |     5.930       5.930
         rho (h/b) |     0.739       0.739
        Unique obs |        11          54

Outcome: dv1. Running variable: running.
--------------------------------------------------------------------------------
            Method |   Coef.    Std. Err.    z     P>|z|    [95% Conf. Interval]
-------------------+------------------------------------------------------------
      Conventional | -.13807     .05677   -2.4318  0.015   -.249342     -.026789
    Bias-corrected | -.30086     .05677   -5.2992  0.000   -.412138     -.189584
            Robust | -.30086      .1136   -2.6484  0.008   -.523518     -.078203
--------------------------------------------------------------------------------
Covariate-adjusted estimates. Additional covariates included: 9
Estimates adjusted for mass points in the running variable.

My question therefore is, how do I manually change the bandwidth for robustness checks, considering I'm interested in the robust options? Should I use the h() or b() option? I'm aware this may be a silly question, but thank you for your help.

Many thanks,
Cat