Description Usage Arguments Details Value
Because the likelihood surfaces for these problems often has a severe bananna shape with a poorly defined maximum finding an optimum is often non-trivial If an optimum is given in the input file I use that as an initial value for nlm otherwise, I found that nlm often gets stuck in a local min. So here I have iterated between a genetic optimization algorithm and nlm. This works more often but still at times misses the optimum. This is adhoc.
The covariates are log-transformed and centered if input$centerMS=TRUE or input$centerFlow=TRUE.
1 | findOptimum(dat, input, silent = FALSE)
|
dat |
data from the A & P file |
input |
a list with the other values needed for a DM run. |
silent |
(TRUE/FALSE) |
In SRFunctions(), bev-holt is defined as S/(S*exp(-p[2])+exp(-p[1]))*exp(p[3]*logMS)*exp(p[4]*logFlow) In DM (writeBUGSmodel.R), R = [S/( (S/exp(logCap)) + (1/prod) )] exp(marineInd*logMS) exp(flowCoef*logFlow) so p[1] = log(prod), constrained to be positive p[2] = log(cap), constrained to be positive p[3] = msCoef, p[4] = flowCoef
A list. $estimate parameters at the minimum sum of squared residuals. The parameters are prod, cap, msCoef, flowCoef. $value is the sum of squared residuals at the minimum.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.