I have a question regarding Optimal Design software.

Particularly, let's suppose I have an intervention that assigns police patrols to different street segments, and that these segments are randomized so that half receive the treatment and the other half doesn't. If I am doing this, should I use person randomized trials, or the would the correct decision be to use clustered randomized trials with clustered-level outcomes?

  • $\begingroup$ Is this really a software question? $\endgroup$
    – Giskard
    Jun 17, 2022 at 12:48
  • 1
    $\begingroup$ Sounds like a cluster-level intervention to me. But another question: Shouldn't we expect contamination if treatment and control are on the same street? Especially if we talk about crime outcomes. $\endgroup$
    – Papayapap
    Jun 17, 2022 at 15:20
  • $\begingroup$ Yes, there could be some contamination, but generally the literature determines a minimum distance of 200m-250m between street segments to prevent that issue. Thank you for solving my question! $\endgroup$
    – Nacho
    Jun 18, 2022 at 2:03


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.