Trying to make sense of a funny result with 'do.fp'

posts related to the RPresence library, which may not be of general interest to users of 'classic' PRESENCE.

Trying to make sense of a funny result with 'do.fp'

Postby cbalantic » Wed May 02, 2018 12:34 pm

Hi there,

I'm running into a funny "edge case" outcome that I'm trying to understand. I'm not sure if this is an Rpresence problem, a Presence problem, or just an issue with design/theory. I'm running simulations using the dynamic model with false positives ('do.fp'), specified as the multiple detection states model (Miller et al. 2013). We're dealing with automated acoustic monitoring data, where some detections can be uncertain (1) because they were only detected automatically, and other detections can be certain (2) if the detected event is later checked and manually confirmed by a human.

We've been looking at a variety of confirmation levels. Generally, as manual confirmation (certain detections) increases, the bias and precision of the state parameter estimates decrease -- as you'd expect. However, I decided to check on what would happen if 100% of uncertain detections were confirmed (resulting in an encounter history consisting strictly of uncertain absences (0) and certain detections (2)). For most of our simulation scenarios, this follows the anticipated trend of decreased bias for all three state parameters (psi, gamma, epsilon), but in a few specific simulation scenarios, we get consistent, substantial overestimates in psi. (Meanwhile, the estimates for the three detection parameters, b, p11, p10, seem to be fine).

I'm trying to understand how this might be. Obviously, if you have 100% confirmation of your surveys, you wouldn't need the false positives model anyway, so I know it's not meant to be used this way, but I'm wondering if there is a clear explanation for what could be happening.

Our simulated truth was psi = 0.6 and epsilon = gamma = 0.25.

As an aside, while I don't get any numerical convergence warnings in the below example, I do get VC warnings and I'm not clear on how to handle those. I'm guessing they are related to the variance-covariance matrix (?). If I were fitting models with covariates, I might have wondered if I have a zero-variance predictor and maybe that was producing the NaNs (?). But since we're fitting intercept models below, I'm not sure what to make of these warnings!

Reproducible example is below... apologies for the long ugly dput() of the encounter.history, but this way you can pop all of the code directly into R and reproduce.

Thank you for reading, and for any insights or explanations you might have for this phenomenon!

- Cathleen Balantic


#### R code to reproduce funky psi overestimate:

library(RPresence)

# Generate example encounter history:
# 100 sites (rows)
# 20 surveys (columns), demonstrating two seasons with 10 surveys each
encounter.history <- structure(list(X1.1 = c(0L, 2L, 0L, 0L, 0L, 2L, 2L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 0L, 2L, 0L, 2L, 2L,
2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L,
2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L,
0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L,
2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L,
2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.2 = c(0L, 2L, 0L,
0L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L,
0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L,
2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L,
0L),
X1.3 = c(0L, 2L, 0L, 0L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L,
0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L,
2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L,
0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L,
2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.4 = c(0L, 2L, 0L, 0L, 0L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L,
0L, 2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L,
2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L,
0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L,
2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.5 = c(0L, 2L, 0L, 0L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L,
2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L,
2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L,
0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.6 = c(0L,
2L, 0L, 0L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L,
2L, 2L, 0L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L,
2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L,
2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L,
0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.7 = c(0L, 2L, 0L,
0L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L,
0L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L,
2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L,
2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L,
0L, 0L, 2L, 2L, 2L, 2L, 0L),
X1.8 = c(0L, 2L, 0L, 0L, 0L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L,
2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L,
2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L,
2L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L,
0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L,
2L, 2L, 2L, 2L, 0L),
X1.9 = c(0L, 2L, 0L, 0L, 0L, 2L, 2L,
0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 0L,
2L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L,
2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L,
2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L,
2L, 2L, 0L),
X1.10 = c(0L, 2L, 0L, 0L, 0L, 2L, 2L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 0L, 2L, 0L,
2L, 2L, 2L, 2L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 0L,
2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 2L, 2L,
2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L,
0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L,
2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L,
0L),
X2.1 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L,
0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L,
0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L,
2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.2 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L, 2L,
2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L,
2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 2L,
2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L, 2L,
0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.3 = c(0L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 0L, 2L,
2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.4 = c(0L, 2L, 0L,
2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L,
2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L,
2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L,
2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L,
2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L,
0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L,
0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.5 = c(0L, 2L, 0L, 2L, 0L,
2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L,
0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L,
0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L,
0L, 0L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L,
0L, 2L, 2L, 2L, 0L),
X2.6 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L,
0L, 0L, 2L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L,
2L, 0L, 2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L,
0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L,
2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L,
0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L,
0L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L,
2L, 2L, 0L),
X2.7 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L,
2L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L,
2L, 2L, 0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L,
0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L,
2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L,
0L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L,
0L),
X2.8 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L,
0L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L,
0L, 2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L,
2L, 2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L,
2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.9 = c(0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L, 2L,
2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L,
2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L,
2L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 2L,
2L, 2L, 2L, 2L, 2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L, 2L,
0L, 0L, 0L, 2L, 2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L),
X2.10 = c(0L,
2L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 0L, 0L,
2L, 2L, 2L, 0L, 2L, 0L, 2L, 0L, 2L, 2L, 0L, 2L, 0L, 0L, 2L,
2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 2L, 2L, 2L, 2L, 0L, 2L, 0L, 0L, 0L,
0L, 2L, 2L, 2L, 0L, 0L, 0L, 0L, 0L, 2L, 2L, 2L, 2L, 2L, 2L,
2L, 0L, 0L, 2L, 0L, 0L, 0L, 2L, 0L, 2L, 2L, 0L, 0L, 0L, 2L,
2L, 2L, 0L, 0L, 0L, 2L, 2L, 2L, 0L)),
.Names = c("X1.1",
"X1.2", "X1.3", "X1.4", "X1.5", "X1.6", "X1.7", "X1.8", "X1.9",
"X1.10", "X2.1", "X2.2", "X2.3", "X2.4", "X2.5", "X2.6", "X2.7",
"X2.8", "X2.9", "X2.10"), class = "data.frame", row.names = c(NA,
-100L))

# Check on output:
# 100 sites (rows)
# 20 surveys (columns), demonstrating two seasons with 10 surveys each
encounter.history

# Create pao to be fed into PRESENCE:
one.pao <- createPao(data = encounter.history,
nsurveyseason = c(10, 10)) # two seasons, 10 surveys each

# Fit model:
m0 <- occMod(model = list(psi~1, gamma~1, epsilon~1, p11~1, p10~1, b~1),
data = one.pao,
type = 'do.fp',
outfile = 'modname')
# note 12 warnings: 12: In sqrt(var) : NaNs produced

# Check on the estimates
est.list <- lapply(m0$real, '[[', 1)
est.list$psi # estimates psi to be 1 at all sites, although it was simulated to be 0.6
est.list$gamma # estimates gamma as 0.237 at all sites (simulated 0.25)
est.list$epsilon # estimates epsilon as 0.306 at all sites (simulated 0.25) (this trends much closer to 0.25 when we perform many replicates of the simulated scenario)
est.list$p10 # estimates approach 0 (as expected, since we have confirmed all surveys)
est.list$p11 # estimates 1 (should be close to 1 since we have confirmed all surveys and simulated a highly detectable species)
est.list$b # estimates 1 (should be close to 1 since we have confirmed all surveys and simulated a highly detectable species)
cbalantic
 
Posts: 5
Joined: Mon Jan 29, 2018 12:20 pm

Re: Trying to make sense of a funny result with 'do.fp'

Postby darryl » Wed May 02, 2018 1:42 pm

Hi Cathleen,
Off the top of my head I'd guess that with no uncertain detections (1's), some of your parameters may not be estimable, or will be estimated right on the boundary of allowable values which can cause the problems with obtaining the variance-covariance matrix your experiencing. This is when you might want to fix some values instead of estimating them.

That doesn't explain your overestimate of psi though...

Cheers
Darryl
darryl
 
Posts: 495
Joined: Thu Jun 12, 2003 3:04 pm
Location: Dunedin, New Zealand

Re: Trying to make sense of a funny result with 'do.fp'

Postby jhines » Wed May 02, 2018 1:46 pm

Hi,

You have 3 estimates (p11,p10,b) which are at the limits (p11=1, p10=0, b=1), which is causing the var-cov warnings. When an estimate is at a boundary (zero or one), it's variance is undefined. Some folks report that as zero, and others report it as N/A.

Also, your starting values are very far from the final values, so PRESENCE is converging on a local minimum neg. log-likelihood. You can specify better starting values using the 'initvals' parameter in the OccMod function. (eg.,

Code: Select all
 #                                     initial values of psi,gam,eps,p11,p10,b
init.vals=qlogis(c(.6,.25,.25,.9,.1,.9)) #   (need to transform w/ qlogis)
m1 <- occMod(model = list(psi~1, gamma~1, epsilon~1, p11~1, p10~1, b~1), data = one.pao,
             initvals=init.vals, type = 'do.fp',   outfile = 'modname')
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Trying to make sense of a funny result with 'do.fp'

Postby cbalantic » Fri May 04, 2018 11:23 am

Specifying more informed inits does the trick -- I'm no longer observing any weird results. Thank you for the boundary conditions insights, as well as the code sample, and thank you both for your quick responses!
cbalantic
 
Posts: 5
Joined: Mon Jan 29, 2018 12:20 pm


Return to RPresence

Who is online

Users browsing this forum: No registered users and 8 guests