抑制来自deSolve :: lsoda的错误

Question

I'm running some algorithms that use approximate Bayesian computation (eg see Toni et al., 2009) and these require repeatedly solving the Lotka-Volterra system of equations with a randomly generated set of input parameters. I'm using the lsoda function from the deSolve package .

Occasionally this function throws an error and I was hoping to ignore this using the try(..., silent = TRUE) function although this doesn't seem to work (see example below). Setting options(show.error.messages = FALSE) doesn't work either.

How do I suppress printing of the error message from deSolve::lsoda?

require(deSolve)

# Differential equations defining the system
LV <- function(Time, State, Pars){

    with(as.list(c(State, Pars)), {

    dx <- a*x - x*y
    dy <- b*x*y - y

    return(list(c(dx, dy)))
    }
    )
}

# Parameters
pars <- c(a = 1.0, b = 1.0)

# Initial conditions
init <- c(x = 1.0, y = 0.5)

# Time steps
times <- seq(0, 15, length.out = 100)

problem_seeds <- c(7, 241, 361, 365, 468, 473, 649, 704, 724, 745, 838)

for (i in problem_seeds){
    set.seed(i)

    # Sample from pi(theta), prior for parameter vector (a, b)
    theta_star <- runif(2, -10, 10)
    names(theta_star) <- c("a", "b")

    # Simulate a dataset using these parameters (at the specified times)
    try(out <- lsoda(func = LV, 
            y = init, 
            parms = theta_star, 
            times = times), 
            silent = TRUE)
    dfs <- as.data.frame(out)
}

Answer 1

Look at page 44 of the vignette in deSolve , this kind of error is described here .

You can solve this problem in reducing the absolute tolerance of the solution. In your example, the following approach works:

out <- lsoda(func = LV, 
             y = init, 
             parms = theta_star, 
             times = times,
             atol = 1e-3)

Note: your data.frame dfs will be overwritten in every loop and if you want the output of the problem_seeds in a data.frame, you can run a fuinction of the apply family. And you don't need the try function from now on.