grayclhn/oosbootstrap

Public repository for my paper on bootstrapping out-of-sample statistics

oosbootstrap's Issues

Allow l/P to converge to a value other than zero

Right now the asymptotic arguments implicitly assume that l/P converges to zero. I don't think that's necessary, so I should relax it.

Lengthen description in Monte Carlo

Edit and proofread Lemma A2

Add discussion of multiple testing

Extend Monte Carlo simulations

The Monte Carlo simulations are limited for a few reasons.

• I just show size results, nothing about power.
• The comparison to other bootstrap approaches is pretty limited. (Admittedly, that's because I'm not aware of many bootstrap results for non-nested models.)
• Just one data generating process, and a somewhat unusual one at that.

So I need to start thinking about how to make progress on that.

Add results and build artifacts to repository

Results in the main branches.
Add a separate public branch for the intermediate but costly build artifacts.

Expand conclusion

Edit to make the paper nicer and more readable

I'm creating this issue to 'collect' the many small edits I expect to make.

Edit/proofread Lemma A1

Edit and proofread Lemma A4

Import project from university GitLab

Edit and proofread Lemma A3

Flesh out Section 2

Too curt

Edit and proofread mathematical appendix

Basic steps:

• Theorem 1, statement and proof
• Theorem 2, add statement and proof
• Lemma A1
• Lemma A2
• Lemma A3
• Lemma A4
• Lemma A5, convergence of bootstrap variance
• Lemma A6, normality of bootstrap terms
• Basic introduction to the appendix

Edit and proofread Lemma A5

Add proof of Theorem 2

Flesh out empirical results