1: \begin{abstract} 

2: Inspired by sample splitting and the reusable holdout introduced in the field

3: of differential privacy, we consider selective inference with a randomized

4: response. We discuss two major advantages of using a randomized response for

5: model selection. First, the selectively valid tests are more powerful after

6: randomized selection. Second, it allows consistent estimation and weak

7: convergence of selective inference procedures.  Under independent sampling, we

8: prove a selective (or privatized) central limit theorem that transfers

9: procedures valid under asymptotic normality without selection to their

10: corresponding selective counterparts. This allows selective inference in

11: nonparametric settings. Finally, we propose a framework of inference after

12: combining multiple randomized selection procedures.  We focus on the classical

13: asymptotic setting, leaving the interesting high-dimensional asymptotic

14: questions for future work.

15: %Using a randomized response can ensure that the

16: %leftover information of \cite{optimal_inference} is bounded below

17: %ensuring that selective intervals are better behaved than without randomization.

18: %Under independent sampling, we prove a selective (or privatized) central limit theorem

19: %that transfers procedures valid under asymptotic normality without selection to 

20: %their corresponding selective counterparts. This allows selective inference in the nonparametric

21: %settings. Finally, we describe 

22: %a method for selective inference following cross-validation using slightly more randomization

23: %than the split into groups of standard cross-validation.

24: %We focus on the classical asymptotic setting, leaving the interesting

25: %high-dimensional asymptotic questions for future work.

26: \end{abstract}