NEWS.md

In mathieucarmassi/calibrationcode: Code Calibration in a Bayesian Framework

CaliCo 0.1.1

Major changes

• To establish a model with a Gaussian process that emulates the code, the user has now three possibilities. The first, he possesses a code written in R but no DOE. The second, he possesses a code written in R and wants to enforce its own DOE. Then, he has no code but possesses the DOE and the corresponding output of a code.
• A pipe is defined to parametrize a statistical model.
• The function plot for parametrized model takes now only two arguments and is simpler than in version 0.0.1
• A fonction forecast replaces the function prediction which did not work very very. The new function forecast allows to predict, based on realized calibration, over a new data set the behaviour of the model.
• Two functions chain and estimators are added to get access to the MCMC chains and the MAP and Mean a posteriori estimator of a calibrated object.
• The ggplots given for the plot of the calibrated object has changed and now to graphics layout are given. The user is also free to load the ggplots into a variable and draw the graph he wants.
• A function that improves calibration for the second and the fourth model have also been written and is called sequentialDesign.

Bug fixes

• The likelihoods written in the models are fixed
• The R6 classes to create the statistical model have been simplified to have a more robust programmation

mathieucarmassi/calibrationcode documentation built on Aug. 14, 2019, 12:35 a.m.