Set class `spp` attributes to mirror MSOM output

[rBatt]

Make it easy to directly compare the simulation results to the MSOM output

Dimensions to consider:

• Reps
• Space
• Time
• Species
• or Aggregated over the above

  Items to test for sensitivity

• Taxonomy/ detectability parameter (p)
• K = number of substrata sampled (n.ss.mu here)
• J = number of grid cells (sites); grid.h and grid.w

  Specific parameters include

• [x] Z in space, time
• [x] Psi in space, time (see here for a start, although not yet perfect)
  • [x] Can use dsample, but can tail off too quickly
  • [x] True value can be computed by dnorm on mus and sds
  • [ ] logistic regression is already used in MSOM
  • [x] tried logistic regression for obs, but inaccurate and throws too many errors/ warnings
  • Potential explanation here (stack overflow), or here (cross validated)
  • Thus, either have to use dsample for obs, or skip
  • [x] p is the probability of detection
  • [x] not current a tracked parameter in MSOM, but u.a0 is tracked
  • [x] can't have "true" and "observed" for p; see here for true values

    Ensure convenience for comparison, and sensitivity

• [ ] Make sure formats from MSOM and spp class equal
  • [x] edit spp class attributes for format and content
  • [x] write functions to convert msom output, or add msom class
• [x] formats should make it easy to associate parameter values with sensitivity values
  • note that detectability is a sensitivity and a parameter!
  • [ ] easiest way is probably to have additional arrays that mimic parameter formats and contain sensitivity values

[rBatt]

Just like it only makes sense to get the probability of detection (p) from the "observation" part of the simulation, it really only makes sense to get the probability of occupancy (Psi) from the "process" (I should start calling this Nature's Data Generating Process, and obs the observation process) part of the simulation. Thus, I needn't worry about how to get Psi from obs.

[rBatt]

I cannot, for the life of me, figure out how the hell to track p as a parameter in JAGS if some of those nodes are missing. At least not in a way that is reasonable.

[rBatt]

I want to do the simulation differently to better emulate how MSOM does its calculations