The power of unbiased recursive partitioning

Blog

news

The significance tests underlying the unbiased tree algorithms CTree, MOB, and GUIDE are embedded into a unifying framework. This allows to assess relative strengths and weaknesses in a variety of setups, highlighting the advantages of score-based tests (as in CTree/MOB) vs. residual-based tests (as in GUIDE). Read more ›

Hybrid Machine Learning Forecasts for the 2019 FIFA Women's World Cup

Blog

news

Using a random forest ensemble learner we obtain probabilistic forecasts for the FIFA Women's World Cup in France: The clear favorite is defending champion USA followed, with some margin, by host France, England, and Germany. Read more ›

Personalized treatment effects with model4you

Blog

news

Personalized treatment effects can be estimated easily with model-based trees and model-based random forests using the R package model4you. Read more ›

At the end of the rainbow

Blog

news

Fully saturated RGB rainbow colors are still widely used in scientific visualizations despite their widely-recognized disadvantages. A recent wild-caught example is presented, showing its limitations along with a better HCL-based alternative palette. Read more ›

Online color apps at hclwizard.org

Blog

news

The hclwizard.org web page has been relaunched, hosting three online color apps based on the HCL (Hue-Chroma-Luminance) color model: a palette constructor, a color vision deficiency emulator, and a color picker. Read more ›

Spatial lag model trees

Blog

news

Economic growth models are recursively partitioned to assess heterogeneity in growth and convergence across EU regions while adjusting for spatial dependencies. Accompanied by R package lagsarlmtree, combining partykit::mob and spdep::lagsarlm. Read more ›

colorspace: New tools for colors and palettes

Blog

news

A major update (version 1.4.0) of the R package colorspace has been released to CRAN, enhancing many of the package's capabilities, e.g., more refined palettes, named palettes, ggplot2 color scales, visualizations for assessing palettes, shiny and Tcl/Tk apps, color vision deficiency emulation, and much more. Read more ›

Minimum CRPS vs. maximum likelihood

Blog

news

In a new paper in Monthly Weather Review, minimum CRPS and maximum likelihood estimation are compared for fitting heteroscedastic (or nonhomogenous) regression models under different response distributions. Minimum CRPS is more robust to distributional misspecification while maximum likelihood is slightly more efficient under correct specification. An R implementation is available in the crch package. Read more ›

Partially additive (generalized) linear model trees

Blog

news

The PALM tree algorithm for partially additive (generalized) linear model trees is introduced along with the R package palmtree. One potential application is modeling of treatment-subgroup interactions while adjusting for global additive effects. Read more ›

Thunderstorm forecasting with GAMs

Blog

news

Boosted binary generalized additive models (GAMs) with stability selection and corresponding MCMC-based credibility intervals are discussed in a new MWR paper as a probabilistic forecasting method for the occurrence of thunderstorms. Read more ›