Reparametrizing Joback's Group Additivity Theory – Thermochemical Building Blocks for Targeted Design of New Compounds

Abstract This study aims to investigate if complex machine‐learning models are required to predict certain properties of materials, or if simple linear regression models also give sufficiently accurate predictions of thermochemical properties and density. An in‐depth analysis of Joback's method in combination with statistical models, and the density predictions by Holden, Keshavarz, and Bondarchuk are made. Updated Group increment tables for Joback's method are provided. The ‘P arametrization of E mpirical A pproximations using S tatistical Methods’ ( PEAS ) is intended to evaluate if it could be better to resort to simpler methods in certain cases. Additionally, a data set analysis tool named “Gr oup E xtraction A nalysis T ool for D eviations in a pproximating n umerical modeling e rrors” ( Great Dane© ) is presented.