The hardware and bandwidth for this mirror is donated by dogado GmbH, the Webhosting and Full Service-Cloud Provider. Check out our Wordpress Tutorial.
If you wish to report a bug, or if you are interested in having us mirror your free-software or open-source project, please feel free to contact us at mirror[@]dogado.de.

midr 0.5.0

First release on CRAN.

midr 0.4.9.909

• Working paper is now available on arXiv.
• Updates for final adjustments.
• Some requirements for the ‘x’ argument are removed or relaxed.

midr 0.4.9.908

• Fix small bugs including one reported in https://github.com/ryo-asashi/midr/issues/5.
• Change the default values of k to avoid singular fits with smaller datasets.

midr 0.4.9.907

• interpret() now includes a new verbosity argument for logging.
• Fix small bugs.

midr 0.4.9.906

• Important: Update the behavior of get.yhat() for classification tasks: If the model returns a matrix or data.frame of class probabilities, by default, get.yhat() returns the probability of not being the base level.

midr 0.4.9.900-0.4.9.905

• ggmid.mid.conditional() and plot.mid.conditional() now include a new argument reference, which allows setting the reference point of c-ICE plot to any of the sample points.
• color.theme() now includes a new argument, pkg, for package specification.
• A new article about “color.theme” objects is added.
• A small change of print.mid() and print.mid.conditional().
• Important: Update and chane the behavior of interpret() for classification tasks: If y is a factor or character, interpret() convert its base level to 0 and all other levels to 1.
• Modify ggmid() and plot.mid() to correct effect plots for factor variables with a catchall level. Additionallym ggmid() now utilize ggplot2::geom_jitter() and allow for adjustable jitter amounts with the jitter argument. Additionally, when data is not explicitly provided, it is now automatically extracted from the function call stored in the "mid" object.
• Adjust mid.conditional() and mid.breakdown() so they no longer require explicit data input.
• Update terminology.
• The color palettes of the khroma package are now available for color.theme().
• The default theme for ggmid(type = "data") and plot(type = "data") is changed to a sequential color scheme: “bluescale”.
• Add link functions: translogit, transprobit, identity-logistic and identity-gaussian for the interpretation task of classification models.
• interpret() now interactively confirms whether a singular fit or exceeding the maximum number of columns is an error.

midr 0.4.3-0.4.8.909

• Add mid.ur() to extract uninterpreted ratio (rate) more conveniently.
• Modify weighted loss functions to compute deviation-from-mean based losses for one input.
• Modify ggmid.mid.breakdown() , ggmid.mid.importance() and plot.mid.breakdown() to improve usability of the functions.
• Modify interpret() to add the pred.args argument that can be used to pass optional arguments to the prediction function (pred.fun()).
• Add a draft of the article on “the interpretation of classification models”.
• Modify interpret() to allow matrices to be used as valid inputs for data (interpret.formula()) and x (interpret.default()).
• Add an article on “the interpretation of regression models”.
• Define print.encoder() for “encoder” objects to improve usability of the fitted MID models.
• Update README.
• Change default values of some arguments: e.g. na.action = na.pass of predict.mid().
• Fix small bugs of plot.mid.conditional().
• Update interpret() to ensure that the value of the argument link is a character string.
• Update ggmid.mid.conditional() and plot.mid.conditional(): var.color and the similar arguments can take an expression as input.

midr 0.4.3

midr is in the release process. We will submit the package to CRAN by mid-January 2025 .

• Improving usability of graphing functions.
• Finalizing code and documentation.
• Writing some vignettes. [Submitted to JARIP BULLETIN]

These binaries (installable software) and packages are in development.
They may not be fully stable and should be used with caution. We make no claims about them.
Health stats visible at Monitor.