calc_MinDose4(input.data, sigmab, log = TRUE, sample.id = "unknown sample",
gamma.xlb = 0.1, gamma.xub = 100, mu.xlb = 1, mu.xub = 100,
sigma.xlb = 0.001, sigma.xub = 5, init.gamma = 10, init.mu = 10,
init.sigma = 0.6, init.p0 = 0.01, ignore.NA = FALSE, calc.ProfileLikelihoods = TRUE,
console.ProfileLikelihoods = FALSE, console.extendedOutput = FALSE,
output.plot = TRUE, output.indices = 4)RLum.Results data.frame: two columns with De (input.data[,1]) and
De error (values[,2])numeric (required): spread in De values given as a
fraction (e.g. 0.2). This value represents the expected overdispersion in
the data should the sample be well-bleached (Cunningham & Walling 2012logical (with default): fit the (un-)logged three parameter
minimum dose model to De datacharacter (with default): sample idnumeric (with default): lower boundary of gammanumeric (with default): upper boundary of gammanumeric (with default): lower boundary of munumeric (with default): upper boundary of munumeric (with default): lower boundary of sigmanumeric (with default): upper boundary of sigmanumeric (with default): starting value of gammanumeric (with default): starting value of munumeric (with default): starting value of sigmanumeric (with default): starting value of p0logical (with default): ignore NA values during log
likelihood calculations. See details.logical (with default): calculate profile log likelihood
functions for gamma, mu, sigma, p0. See output.indices.logical (with default): print profile log likelihood
functions for gamma, mu, sigma, p0 to console.logical (with default): extended terminal outputlogical (with default): plot output
(TRUE/FALSE)numeric (with default): requires
calc.ProfileLikelihoods = TRUE. Indices: 1 = gamma,
2 = gamma/mu, 3 = gamma/mu/sigma, 4 = gamma/mu/sigma/p0RLum.Results get_RLum.Resultsgamma: minimum dose on the log scale
mu: mean of the non-truncated normal distribution
sigma: spread in ages above the minimum
p0: proportion of grains at gamma
}
(Un-)logged model
In the original version of the three-parameter minimum dose model, the
basic data are the natural logarithms of the De estimates and relative
standard errors of the De estimates. This model will be applied if
log = TRUE.
If log = FALSE, the modified un-logged model will be applied
instead. This has essentially the same form as the original version.
gamma and sigma are in Gy and gamma becomes the
minimum true dose in the population.
While the original (logged) version of the mimimum dose model may be
appropriate for most samples (i.e. De distributions), the modified
(un-logged) version is specially designed for modern-age and young
samples containing negative, zero or near-zero De estimates (Arnold
et al. 2009, p. 323).
Boundaries
Depending on the data, the upper and lower bounds for gamma
(gamma.xlb and gamma.xub) and mu (mu.xlb and
mu.xub) need to be specified. If the final estimate of gamma or
mu is on the boundary, gamma.xlb and gamma.xub
(mu.xlb and mu.xub respectively) need to be adjusted
appropriately, so that gamma and mu lie within the bounds. The same
applies for sigma boundaries (sigma.xlb and sigma.xub)
Initial values
The log likelihood calculations use the nlminb function.
Accordingly, initial values for the four parameters init.gamma,
init.sigma, init.mu and init.p0 need to be specified.
Ignore NA values
In some cases during the calculation of the log likelihoods NA values
are produced instantly terminating the minimum age model. It is advised to
adjust some of the values provided for any argument. If the model still
produces NA values it is possible to omit these values by setting
ignore.NA = TRUE. While the model is then usually able to finish
all calculations the integrity of the final estimates cannot be ensured.
Use this argument at own risk.nlminb,
calc_CentralDose,
calc_CommonDose, calc_FiniteMixture,
calc_FuchsLang2001, calc_MinDose3## load example data
data(ExampleData.DeValues, envir = environment())
## apply the logged minimum dose model
calc_MinDose4(ExampleData.DeValues,
sigmab = 0.05, gamma.xub = 10000, mu.xub = 10000, init.p0 = 0.4,
output.plot = FALSE)Run the code above in your browser using DataLab