Functions available to any gadget3 model

Details

g3_env is the top-level environment that any gadget3 model uses, populated with utility functions.

NB: Several functions have _vec variants. Due to TMB limitations these should be used when you have a vector not scalar input.

ADREPORT

TMB's ADREPORT function. See sdreport documentation

as_integer

C++ compatible equivalent to as.integer

as.numeric

R as.numeric or TMB asDouble

assert_msg

C++/R function that ensures expression is true, or stops model.

assert_msg(x > 0, "x must be positive")

avoid_zero / avoid_zero_vec

Adds small value to input to ensure output is never zero

bounded / bounded_vec

Ensures x is within limits a & b.

bounded_vec(x, 100, 1000)

g3_matrix_vec

Apply matrix transformation tf to vector vec, return resultant vector.

g3_matrix_vec(tf, vec)

lgamma_vec

Vector equivalent of lgamma

logspace_add / logspace_add_vec

TMB's logspace_add, essentially a differentiable version of pmax.

normalize_vec

Divide vector a by it's sum, i.e. so it now sums to 1

nvl

Return first non-null argument. NB: No C++ implementation.

pow_vec

Vector equivalent of ^

ratio_add_vec

Sum orig_vec & new_vec according to ratio of orig_amount & new_amount

REPORT

TMB's REPORT function.

REprintf

Equivalent of RCpp REprintf

Rprintf

Equivalent of RCpp Rprintf

Examples

## avoid_zero / avoid_zero_vec
g3_eval(quote( c( avoid_zero(0), avoid_zero(10) ) ))
#> [1] 6.931472e-04 1.000000e+01
g3_eval(quote( avoid_zero_vec(0:5) ))
#> [1] 0.0006931472 1.0000000000 2.0000000000 3.0000000000 4.0000000000
#> [6] 5.0000000000

## bounded / bounded_vec
curve(g3_eval(quote( bounded(x, 100, 200) ), x = x), -100, 100)


## logspace_add
curve(g3_eval(quote( logspace_add(x, 10) ), x = x), 0, 40)


## normalize_vec
g3_eval(quote( normalize_vec(c( 4, 4, 8, 2 )) ))
#> [1] 0.2222222 0.2222222 0.4444444 0.1111111