search

pfmc-assessments / VASTWestCoast

VAST for the NWFSC West Coast data
Other
5 stars 1 forks source link

research: regression discontinuity simulation #31

Open kellijohnson-NOAA opened 4 years ago

kellijohnson-NOAA commented 4 years ago

@James-Thorson-NOAA suggested doing a simulation to address Regression Discontinuity in #26. Some of the pertinent text from #26 is quoted below, but I wanted to start a new thread in case someone is interested in working on it.

... help completing a paper doing a simulation study showing the performance of regression discontinuity designs if anyone there has time and interest. I agree it's a good path towards integrating BC and US surveys :)

On Wed, Sep 2, 2020, 8:36 PM mhaltuch notifications@github.com wrote: This method is also what Maia used to create an index of abundance, using WC, BC, and AK surveys, for the NE Pacific sablefish OM.

On Wed, Sep 2, 2020 at 12:19 PM Jim Thorson notifications@github.com wrote: We have submitted a paper doing this for arrowtooth using the triennial and annual bottom trawls plus BC and AFSC surveys. It's mainly focused on a global synthesis of bottom trawls survey extent. I've been calling "calibration by proximity" a regression discontinuity design to link to the stats literature on the topic. I'd be happy to chat more about it offline. Main surveys seem like

  1. Bottom trawls
  2. Hook and line
  3. Accoustics for hake and sardine
  4. Calcofi
  5. Juvenile rockfish
  6. Newport line I'd of course be happy to help with synthesis for any or all of those! Or fine to not be involved too

On Wed, Sep 2, 2020, 11:35 AM Ian Taylor notifications@github.com wrote: Years ago @James-Thorson https://github.com/James-Thorson and Robyn Forrest were working on a project looking across regions which Jim described as follows: integrating bottom-trawl data from non-overlapping designs, e.g.: California Current (triennial + shelf-slope), BC (many programs), Gulf of Alaska, and Eastern Bering Sea using what I call "calibration by proximity" (the assumption that survey boundaries are statistically independent of underlying biological variation) but I don't know if it went anywhere (I don't see anything on Google Scholar).

Originally posted by @James-Thorson in https://github.com/nwfsc-assess/VASTWestCoast/issues/26#issuecomment-686237465

kellijohnson-NOAA commented 4 years ago

Thanks @James-Thorson for mentioning the need for simulation studies on this topic. I thought this warranted its own issue given the high level of importance here. I would love to work on it, but I have zero additional time until summer 2021. Maybe someone else could take the lead. I am sure that there are multiple frameworks out there that are currently coded that could simulate spatially explicit data like the hake OM. Do you have a simulator in mind? I am much better at organizing EMs than generating data.