pgen: Probability of a Management Parameter Exceeding a Reference...

In fishmethods: Fishery Science Methods and Models

Probability of a Management Parameter Exceeding a Reference Point

Description

Calculates the probability of a management value exceeding a reference point with or without error

Usage

  pgen(est=NULL,limit=NULL,estSD=0,limSD=0,corr=0,dist=1,comp=1,nreps=10000)

Arguments

est	management value (mv) or vector containing individual parameter values from, say, bootstrap runs.
limit	reference point (rp) or vector containing individual reference point values from, say, bootstrap runs.
estSD	standard deviation of management value if a single value is used. Must be >0 if a single value is used. If a vector of individual values is provided, estSD is not used.
limSD	standard deviation of reference point if a single value is used. If a vector of individual values is provided, limSD is not used. limSD = 0 if the reference point is considered a point estimate (no error).
corr	correlation between est and limit. Only used if est and limit are single values with error.
dist	assumed distribution of est or limit if they are single values with error. 1 = normal; 2 = log-normal.
comp	the direction of comparison: 1: mv < rp, 2: mv <= rp, 3: mv > rp, 4: mv >= rp.
nreps	the number of samples to draw to create normal or log-normal distributions. User should explore different sample sizes to determine if the probability obtained is stable.

Details

Randomization methods as approximations to Equations 1, 2 and 3 in Shertzer et al. (2008) are used to calculate the probability that a management value with error (e.g., fishing mortality) passes a reference point without (Eq. 1) or with (Eq. 2) error. Either may be represented by a single value and its associated standard deviations or a vector of individual values that represent results from, say, bootstrap runs. If log-normal is assumed, mv and rp and associated standard deviations must be in natural log-units (i.e., meanlog and sdlog).

If the management value and reference point are specified as single values with standard deviations, samples of size nreps are drawn randomly from the specified distribution parameterized with est and limit and associated standard deviations. If corr>0 (Eq. 3), then the est and limit distributions are drawn from a multivariate normal (function mvrnorm) distribution. If log-normal is assumed, function mvrnorm is used with the meanlog and sdlog estimates and then output values are bias-corrected and back-transformed.

If the management value and the reference point are represented by vectors of individual values, the probability is calculated by tallying the number of management values that exceed (or pass) the reference points and then dividing by number of est values*number of limit values. If either the management value or reference point is specified as a single value with standard deviation, then a vector of individual values of size equal to the size of the other vector is generated by using the rnorm or rlnorm function parameterized with the single value and its standard deviation.

Value

probability value of comparison

Note

Chris Legault of the National Marine Fisheries Service, Woods Hole, MA provided R code for the randomization method and Daniel Hennen of the the National Marine Fisheries Service, Woods Hole, MA provided the R code for using mvrnorm to obtain log-normal distributions.

Author(s)

Gary A. Nelson, Massachusetts Division of Marine Fisheries [e-mail]

References

Shertzer, K. W., M. H. Prager, and E. K. Williams. 2008. A probability-based approach to setting annual catch levels. Fishery Bulletion 106: 225-232.

Examples

 ## est = 2010 Spawning Stock Biomass of Striped Bass, limit = SSB Reference Point
  pgen(est=50548,limit=36881,estSD=5485,limSD=1901,corr=0.05,dist=1,comp=2,nreps=1000)

fishmethods documentation built on April 27, 2023, 9:10 a.m.