NEWS for the R Package RLumModel
Changes in version 0.2.10 (2022-02-20)
This package version requires R >= 4.0
New functions
trace_ParameterStateEvolution()
: this new function allows tracing of the evolution of the concentration during the modelling for different steps. The functions returns a plot and/or numerical output of calculated end-states.
New models
- Add support for Bailey (2001) model variation of Peng et al. (2022, QG), keyword in
model_LuminescenceSignals()
is "Peng2022"
Bugfixes/changes
.set_pars()
- now consistently supports both spellings for the model
"customized"
and "customized"
- now returns a character vector with the names of the allowed keywords if model is missing. This is also internally used in
model_LuminescenceSignals()
to avoid inconsistencies.
- now returns also the parameter names and units as extra elements
model_LuminescenceSignals()
- The parameter
own_state_parameters
now accepts (optionally) an RLum.Results-class
object created by RLumModel::.set_pars
as input
Miscalaneous
- A new vignette, called “RLumModel - Models and parameters” was added to provide a better overview and comparison of the different models