grayclhn/boot-rd

Repository for the paper "Bootstrap Confidence Intervals for Sharp Regression Discontinuity Designs with the Uniform Kernel" (Bartalotti, Calhoun, and He)

boot-rd's Issues

Discovered a small error at the end of Theorem 2's proof.

Fix is coming soon.

Typo in definition of epsilon_{+,i} in Theorem 2.

We've written epsilon_{i,+} instead, which is wrong.

Need to add heteroskedastic DGPs to simulations

Should be fairly straightforward --- Matias suggested looking at MacKinnon's paper for some ideas:
mackinnon_robust.pdf

Basic plan:

• See what Matias has implemented in his recent research.
• Program it up.
• Add wild bootstrap as well
• Run everything.
• Edit text to reflect new results.

[Ref 1.4] Define HC3 transformation

From the report

We need to thank the referees, Matias, Juan Carlos, and other seminar participants.

Is there anyone else that we should specifically thank?

Need to replace residual bootstrap with Wild bootstrap

Main steps (in quasi-order)

• Rewrite algorithm descriptions
• Change proof of Theorem 1
• Change proof of Theorem 2
• Change bootstrap used in simulations
• Change bootstrap used in empirical application
• Edit text of introduction
• Review and potentially edit "background" section
• Tweak conclusion

We should also look at relaxing the assumption on the kernels (which the referees wanted too) while we're doing the wild bootstrap.

Replace g_+ and g_- notation in defining \tau^*

We need to define notation for the bias estimator, more specifically for the value of $\tau{*}{\nu}$, the notation using $g{+}(0)$ and $g_{-}(0)$ is not the easiest to adapt. This will requires changes in the way the algorithm is described as well.

[Ref 1.1 and 1.2] Address similarity between wild bootstrap and CCT's correction

Referee 1's first comment:

His or her second comment is about the same point as well:

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

[Ref 2.1] Address the bandwidth issues raised by the 2nd referee

'typo' is a very charitable way of putting it :)

Add a reference in the introduction alerting the reader about the higher-order refinements of the RDROBUST

discuss and cite the following paper:

Calonico, Cattaneo and Farrell (2016): On the Effect of Bias Estimation on Coverage Accuracy in Nonparametric Inference, working paper, University of Michigan.

Basically one sentence at the end of the second paragraph on p. 3, right after the sentence that reads "The resulting confidence intervals have accurate coverage even when the naive RD interval does not."

Move referee reports into a separate subdirectory

Update citation for RD robust

Could you please replace your citation to the RDROBUST package ("rdrobust: Robust Data-Driven Statistical Inference in Regression-Discontinuity Designs. R package version 0.80") by the following two references?

Calonico, Cattaneo and Titiunik (2015): rdrobust: An R Package for Robust Nonparametric Inference in Regression-Discontinuity Designs, R Journal 7(1): 38-51.

Calonico, Cattaneo, Farrell and Titiunik (2016): rdrobust: Software for Regression Discontinuity Designs, working paper, University of Michigan.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

[Ref 1.3] Make small notation change in Assumption 1

From the report:

HC1 residual adjustment is incorrect

This is not urgent since we don't use them, but we should fix it.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.

[Ref 2.2] Change sup_{x,\tau} to sup_x for referee

Not a big change, but a simple enough request to accommodate.

We're currently reporting the negative bias in simulations, not the bias

See the following lines where we define the bias in the simulations:

boot-rd/R/simulation.R

Line 65 in 6316f92

bias <- t.true - mean(collect.simu[ , 1])

,

boot-rd/R/simulation.R

Line 103 in 6316f92

bias <- t.true - mean(collect.simu[ , 1])

,

boot-rd/R/simulation.R

Line 137 in 6316f92

bias <- t.true - mean(collect.simu[ , 1])

,
and

boot-rd/R/simulation.R

Line 175 in 6316f92

bias <- t.true - mean(collect.simu[ , 1])

They're reporting t.true - mean(simulated values), but the bias is mean(simulated values) - t.true.

Typo at the beginning of Theorem 2: tau-hat(h)* should be tau-hat*(h)

[Ref 2.3] Address variance issues in Theorem 2

From referee 2:

[Ref 1.4] Define beta hat's components earlier

Again, from the report

small change and easy to do.

Describe the distribution for the wild shocks used in the simulation section.

Increment issue counter

I've imported the project to GitHub from GitLab. To make sure that we don't duplicate issue numbers, I'm adding a few empty issues.