analyse_IRSAR.RF(object, sequence.structure = c("NATURAL", "REGENERATED"),
method = "FIT", fit.range.min, fit.range.max, fit.trace = FALSE,
fit.MC.runs = 10, slide.outlier.rm = FALSE, slide.trend.corr = FALSE,
plot = TRUE, xlab.unit = "s", legend.pos, ...)RLum.Analysis character (with default): setting method applied for the data analysis.
Possible options are "FIT" or "SLIDE".integer (optional): set the minimum channel range for signal fitting and sliding.
Usually the entire data set is used for curve fitting, but there might be
reasons to limit the channels used for fittiinteger (optional): set maximum channel range for signal fitting and sliding.
Usually the entire data set is used for curve fitting, but there might be
reasons to limit the channels used for fitting.logical (with default): trace fitting (for debugging use)numeric (with default): set number of Monte Carlo runs for start
parameter estimation. Note: Higher values will significantly increase
the calculation time.logical (with default): enable or disable outlier removal.
Outliers are removed from the natural signal curve only.logical (with default): enable or disable trend correction.
If TRUE, the sliding is applied to a previously trend corrected data set.logical (with default): plot output (TRUE or FALSE)character (with default): set unit for x-axischaracter (with default): useful keywords are bottomright,
bottom, bottomleft, left, topleft,
top, topright, main, xlab, ylab, xlim, ylim,
logRLum.Results data.frame containing De-values with error
(gray dashed lines in the plot) and further parameters. Corrected De values are only
provided for the method "SLIDE", provided the trend correction is applied.nlsModel objectDe.values) should be accessed using the
function get_RLum.Resultsport algorithm of the nls function.
Two methods are supported to obtain the De:
method = "FIT"
The principle is described above and follows the orignal suggestions from Erfurt et al., 2003.
method = "SLIDE"
For this method the natural curve is slided along the x-axis until congruence with the
regenerated curve is reached. Instead of fitting this allows to work with the original data
without the need of any physical model.
This approach was introduced for RF curves by Buylaert et al., 2012 and
Lapp et al., 2012.
Here the sliding is done by searching for the minimum of the residual squares.
$$min(\Sigma(RF.reg_{k.i} - RF.nat_{k.i})^2)$$
for
$$k = {t.0+i,...,t.max+i}$$
Correction for outliers (slide.outlier.rm = TRUE)
By using method = "SLIDE" and setting the argument slide.outlier.rm = TRUE
an automatic outlier removal can be applied to the natural curve. Outliers may be observed also
on the regenerative curve, but here the impact of single outliers on the curve adjustment (sliding)
is considered as negligible.
The applied outlier removal algorithm consists of three steps:
(a) Input data are smoothed using the function rollmedian. Value k for
the rolling window is fixed to 11. Therefore, the natural curve needs to comprise at least of 33
values, otherwise outlier removal is rejected.
(b) To subsequently remove outliers, code blocks from the function apply_CosmicRayRemoval
were recycled, therefore in general the outlier correction works as described by Pych (2003).
In contrast, here no sigma clipping before constructing the histograms is applied.
(c) Outliers are marked in the data set and visualised in the graphical output. The subsequent
adjustement of both curves (natural and regenerative) is done without outliers, whereas the
sliding itself is done with the entire data set.
Trend correction (slide.trend.corr = TRUE)
This option allows for correcting any linear trend in the natural curve in comparison to the
regenerative curve. The trend correction is based on regression analysis of the residuals from
the slided curve. The corrected De is obtained by sliding the trend corrected values (again)
along the regenerative data curve. This correction is driven by the idea that the
rediduals from the regenerative and the natural curve should be free of any trend, as long as
they are comparable.
Error estimation
For method = "FIT" the asymmetric error range is taken from the standard deviation
of the natural signal.
For method = "SLIDE" so far no error estimation is implemented. Instead, to asses
the error of the De several aliquots should be measured and the error obtained from the
De distribution.RLum.Analysis RLum.Results get_RLum.Results, nls##load data
data(ExampleData.RLum.Analysis, envir = environment())
##perform analysis
temp <- analyse_IRSAR.RF(object = IRSAR.RF.Data)Run the code above in your browser using DataLab