devRateModel {devRate}

Compute non-linear regression

Description

Determine the nonlinear least-squares estimates of the parameters of a nonlinear model, on the basis of the nls function from package stats.

Usage

devRateModel(eq, temp, devRate, startValues, dfData = NULL, algo = "GN", ...)

Arguments

eq

The name of the equation. See devRateEqList for the list of equations

temp

The temperature (vector).

devRate

The development rate (days)^-1 (vector).

startValues

Starting values for the regression (list).

dfData

A data.frame with the temperature in the first column and the development rate in the second column (alternative to the use of temp and devRate).

algo

The abbreviated name of the algorithm used for model fitting ( "GN" for Gauss-Newton algorithm, "LM" for Levenberg-Marquardt algorithm; "GN" is the default value).

...

Additional arguments for the nls function.

Details

startValues for equations by Stinner et al. 1974 and Lamb 1992 are composed of two equations: one for the temperatures below the optimal temperature and another for the temperatures above the optimal temperature. For these equations, startValues should be a list of two lists, where the second element only contain starting estimates not specified in the first element, e.g., for Stinner et al.: startValues <- list(list(C = 0.05, k1 = 5, k2 = -0.3), list(Topt = 30)), and for Lamb 1992: startValues <- list(list(Rm = 0.05, Tmax = 35, To = 15), list(T1 = 4))

The temperature should be provided as a vector in argument temp and development rate in another vector in argument devRate. However, it is possible to use the function with a data.frame containing the temperature in the first column and the development rate in the sceond column, using the argument dfData

NULL is returned when an unknown algorithm is entered.

Value

An object of class nls (except for Stinner et al. 1974 and Lamb 1992 where the function returns a list of two objects of class nls).

Examples

## Example with a linear model (no starting estimates)
myT <- 5:15
myDev <- -0.05 + rnorm(n = length(myT), mean = myT, sd = 1) * 0.01
myNLS <- devRateModel(
  eq = campbell_74,
  temp = myT,
  devRate = myDev)
## Example with a non-linear model (starting estimates)
myT <- seq(from = 0, to = 50, by = 10)
myDev <- c(0.001, 0.008, 0.02, 0.03, 0.018, 0.004)
myNLS <- devRateModel(
  eq = stinner_74,
  temp = myT,
  devRate = myDev,
  startValues = list(
    list(C = 0.05, k1 = 5, k2 = -0.3),
    list(Topt = 30)))
## Example with a data.frame instead of two vectors for temperature and
## development rate
myDF <- data.frame(myT, myDev)
myNLS <- devRateModel(
  eq = campbell_74,
  dfData = myDF)