Title: Statistical modelling
Authors : Dettling, Marcel
Ruckstuhl, Andreas
et. al : No
Published in : Applied data science : lessons learned for the data-driven business
Pages : 181
Pages to: 203
Editors of the parent work: Braschler, Martin
Stadelmann, Thilo
Stockinger, Kurt
Publisher / Ed. Institution : Springer
Publisher / Ed. Institution: Cham
Issue Date: 2019
License (according to publishing contract) : Licence according to publishing contract
Type of review: Editorial review
Language : English
Subjects : Data science; Statistical modelling; Regression analysis
Subject (DDC) : 003: Systems
500: Natural sciences and mathematics
Abstract: In this chapter, we present statistical modelling approaches for predictive tasks in business and science. Most prominent is the ubiquitous multiple linear regression approach where coefficients are estimated using the ordinary least squares algorithm. There are many derivations and generalizations of that technique. In the form of logistic regression, it has been adapted to cope with binary classification problems. Various statistical survival models allow for modelling of time-to-event data. We will detail the many benefits and a few pitfalls of these techniques based on real-world examples. A primary focus will be on pointing out the added value that these statistical modelling tools yield over more black box-type machine-learning algorithms. In our opinion, the added value predominantly stems from the often much easier interpretation of the model, the availability of tools that pin down the influence of the predictor variables in concise form, and finally from the options they provide for variable selection and residual analysis, allowing for user-friendly model development, refinement, and improvement.
Departement: School of Engineering
Organisational Unit: Institute of Data Analysis and Process Design (IDP)
Publication type: Book part
DOI : 10.1007/978-3-030-11821-1_11
ISBN: 978-3-030-11820-4
978-3-030-11821-1
URI: https://digitalcollection.zhaw.ch/handle/11475/18174
Appears in Collections:Publikationen School of Engineering

Files in This Item:
There are no files associated with this item.


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.