The new version includes the following updates and improvements:
Restriction of the search region of lambda to the range 0.0 to 1.0 (before it was 0.0 to 1.5)
Slight modification of model estimation for small sample sizes (N < 1000)
Performance improvements, especially for data sets with small sample sizes (N < 1000)
RIbench benchmark score: 0.300, Failure rate: 0.0, Implausible Results: 0.069%
The new version includes the following updates and improvements:
Revised vignette and literature information
Slight modification of search region for the fraction of non-pathological samples
RIbench benchmark score: 0.307, Failure rate: 0.0, Implausible Results: 0.122%
The new version includes the following updates and improvements:
Using the truncated normal distribution now as basic model assumption
Additional regularization term in cost function to control the fit at the tails of the distribution
The median bootstrapping model (pointEst="medianBS") now corresponds to one parametric model selected from all bootstrapping models
Some modifications in plot function, e.g. adjusted default xlim
Slight modification of print function
RIbench benchmark score: 0.307, Failure rate: 0.0, Implausible Results: 0.122%
The new version includes the following updates and improvements:
Adjusted computation of the roundingBase to now also handle data not rounded to power of 10 (e.g. data that was rounded prior to unit conversion)
Vignette ('refineR_package') demonstrating the main functions of the package
RIbench benchmark score: 0.307, Failure rate: 0.0, Implausible Results: 0.625%
The new version includes the following updates and improvements:
More fine-grained search region for lambda (lambdaVec)
Option to use the two-parameter (modified) Box-Cox transformation
(findRI(Data = Data, model = "modBoxCox"))
Update of calculation of costs: new factor to account for small deviations from the assumption of a unimodal distribution of non-pathological samples
Adapted definition of region of test results that characterizes the non-pathological distribution
Improved performance for skewed distributions
RIbench benchmark score: 0.307, Failure rate: 0.0, Implausible Results: 0.625%
Detailed description can be found in Ammer, T., Ammer, T.,Schuetzenmeister, A., Prokosch, HU., Zierk, J., Rank, C.M., Rauh, M. RIbench: A Proposed Benchmark for the Standardized Evaluation of Indirect Methods for Reference Interval Estimation. Clinical Chemistry (2022)
Initial version put on CRAN
Detailed description of the method can be found in Ammer, T., Schuetzenmeister, A., Prokosch, HU., Rauh, M., Rank, C.M., Zierk, J. refineR: A Novel Algorithm for Reference Interval Estimation from Real-World Data. Sci Rep 11, 16023 (2021).