Version 1.4.4
-------------
- bug fixed that multiple fitted models with identical AIC lead to multiple formulas
  and fit statistics added to the plot
- function 'get_model' added implementing the selection of the best model in case
  of identical AIC for multiple models, sequence of complexity can be adjusted
  in function 'fit_vfp' using parameter 'cpx'
- in function 'predict_mean' algorithm to determine intersection refined
- in 'predict_mean' when specified Y-value for which X-value shall be determined
  does not intersect with variance function NA is returned instead of MIN or MAX 
  X-Value to better reflect non-convergence of the algorithm
- unit-test TC15.predictMean adapted to better quality results and to NA instead
  of max X-value in case of no intersection of variance function with Y-value
- bug fixed in 'signif2' returning x if this is among -Inf, NA, NaN, Inf 
- residuals method added for objects of class 'VFP'


Version 1.4.3
-------------
- code was refactored to comply with tidyverse style guide (not 100 percent)
- any function names including '.' were renamed; to be backward compatible
  old function names were made available in .onLoad in .GlobalEnv


Version 1.4.2
-------------
- fixing bug that model 6 as the best fitting model cannot be automatically selected
  issuing an error
- default-setting for 'model.no' changed to 1:9 not taking model 10 into account as it
  does not fit into the class of generalized non-linear models and should not be compared
  to other models 1-9, it is still available


Version 1.4.1
-------------

- PDF-vignette 'How_to_work_with_package_VFP.pdf' was deleted as CRAN-checks
  on linux OS generated warnings complaining about file size


Version 1.4
-------------

- corrected model 2 formula
- specifying 'xlim' in function 'plot.VFP' lead to an error when calling it from
  within another function, this was corrected by adding 'xlim' as function argument
- in function 'precisionPlot', the label at the mean of the respective normal distribution
  was not correct in case of more than two specified probabilities, this was corrected
- degrees of freedom for non-linear models corrected 
- Function 'fit.vfp' was adapted to cases where a valid set of parameters in the re-parameterized
  space was found but the final gnm-fit in the original parameter-space did not converge. In these
  cases the final set of parameters is returned but no variance-covariance matrix is available. 
  Thus, there will not be any confidence-intervals downstream.
- Adaption in the calculation of confidence intervals. These are now always calculated on the vc scale
  and then eventually transformed to the sd or cv scale. 



Version 1.3
-----------

- issue was fixed if input data contained variables with default names but user specified
  different variables, this had no effect because the variables with default names were used
  for the fitting
- condition handler was updated indicating an error if no result is returned by
  package gnm
- multiple issues were reported for model three which did not return a valid result,
  start value procedure for this model was improved 

Version 1.2
-----------

- a package-vignette was integrated 
- function 'precisionPlot' was added visualizing the relationship between
  C5 and C95 concentrations, (im)precision and one or two cutoffs

Version 1.1
-----------

- fixed issue with 'alpha' not having any effect when being set different from the default in function 'plot.VFP'
- added function 'deriveCx' for derivation of sample-concentration at which a pre-defined proportion of repeated
  measurements would fall above the cutoff, e.g. C5 or C95

Version 1.0
-----------

- initial version put on CRAN
