obsfun0 {pttstability}

default observation noise function

Description

Two options: If inverse=FALSE, calculates the log probability density of observation yt based on true state xt and observation error. Otherwise, simulates N random observations of yt. Observation error follows a Gaussian distribution truncated at zero, using a Tobit distribution. Note that probability density is calculated based on a Tobit distribution, with lower boundary zero.

Usage

obsfun0(
  so,
  yt,
  xt = NULL,
  inverse = FALSE,
  N = NULL,
  minsd = 0.01,
  time = NULL
)

Arguments

so

a numeric vector of length one, specifying either log-transformed standard deviation of the observation error as a fraction of the observation, or two log-transformed parameters of the form sd=exp(B0)+exp(B1)*x.

yt

a number, representing a potential observed value of xt

xt

a number or numeric vector of "true" (or simulated) abundances at time t, from which the likelihood of yt will be calculated - defaults to NULL for inverse=TRUE

inverse

a logical specifying whether inverse (i.e. random number generator) function should be implemented - defaults to FALSE

N

number of draws from the random number generator, if inverse=TRUE - defaults to NULL

minsd

minimum observation error allowed (e.g. if observation = 0), to prevent log likelihoods of -infinity - defaults to 0.01

time

the timestep - defaults to NULL (i.e. not used)

Value

If inverse=FALSE, returns a list including LL, a number or numeric vector of length xt, with predicted log likelihoods of observation yt, and wts, a number or vector with weights corresponding to the relative likelihood of each observation (after accounting for variable continuous vs. discrete probability distributions). If inverse = FALSE, returns N random draws from the observation function.