Param_Fn: Data inputs for TMB

In aaronmberger/Geo_dGLMM_habitat: Package for conducting spatial estimation of CPUE index standardization (fisheries, ecology, etc).

Description

Generates a tagged list representing starting values for coefficients (fixed and random effects) estimated by TMB

Usage

Param_Fn(DataList)

Arguments

DataList

Tagged list generated by DataFn()

Details

This function tries to generate logical starting values. For some problems, or for speed, you might start from better-informed values.

Value

This function generates a tagged list, where each slot corresponds to starting values for the following coefficients:

beta1_t	Scaling encounter probability up or down for each year
gamma1_j	Effect of covariate j on encounter probability
logetaE1	Magnitude of marginal variance in spatial variation in encounter probability
logetaO1	Magnitude of marginal variance in spatiotemporal variation in encounter probability
logkappa1	Distance for correlation in encounter probability
logsigmaV1	standard deviation of vessel effects on encounter probability
logsigmaVT1	standard deviation of vessel-year effects on encounter probability
nu1_v	Vessel effects for encounter probability
nu1_vt	Vessel-year effects for encounter probability
Omegainput1_s	Residual spatial variation in encounter probability that is constant for all years
Epsiloninput1_st	Residual spatial variation in encounter probability that varies among year

The preceeding parameters are then repreated for positive catch rates, e.g., beta2_t scales positive catch rates up or down for each year. There are also the following parameters:

ln_H_input	Parameters representing geometric anisotropy
logSigmaM	Parameters representing the distribution of residual errors

Many coefficients are either random (and hence estimated during innner optimization) or 'turned-off' using the tagged list generated by Map_Fn

Examples

## The function is currently defined as
function (DataList) 
{
    Return = list(ln_H_input = c(0, 0), beta1_t = qlogis(tapply(ifelse(DataList$b_i > 
        0, 1, 0), INDEX = DataList$t_i, FUN = mean)), gamma1_j = rep(0, 
        DataList$n_j), lambda1_k = rep(0, DataList$n_k), logetaE1 = 0, 
        logetaO1 = 0, logkappa1 = 0, logsigmaV1 = log(1), logsigmaVT1 = log(1), 
        nu1_v = rep(0, DataList$n_v), nu1_vt = matrix(0, nrow = DataList$n_v, 
            ncol = DataList$n_t), Omegainput1_s = rep(0, DataList$n_s), 
        Epsiloninput1_st = matrix(0, nrow = DataList$n_s, ncol = DataList$n_t), 
        beta2_t = log(tapply(ifelse(DataList$b_i > 0, DataList$b_i/DataList$a_i, 
            NA), INDEX = DataList$t_i, FUN = mean, na.rm = TRUE)), 
        gamma2_j = rep(0, DataList$n_j), lambda2_k = rep(0, DataList$n_k), 
        logetaE2 = 0, logetaO2 = 0, logkappa2 = 0, logsigmaV2 = log(1), 
        logsigmaVT2 = log(1), logSigmaM = c(log(5), qlogis(0.8), 
            log(2), log(5)), nu2_v = rep(0, DataList$n_v), nu2_vt = matrix(0, 
            nrow = DataList$n_v, ncol = DataList$n_t), Omegainput2_s = rep(0, 
            DataList$n_s), Epsiloninput2_st = matrix(0, nrow = DataList$n_s, 
            ncol = DataList$n_t))
    return(Return)
  }

aaronmberger/Geo_dGLMM_habitat documentation built on May 10, 2019, 3:20 a.m.