Curs: "Statistical Modelling for Scientists and Engineers"

El prof David Meintrup de la University of Applied Sciences Ingolstadt impartirà el curs "Statistical Modelling for Scientists and Engineers", els dies 17 i 18 de maig de 12:00 a 14:00. 

INSCRIPCIÓ TANCADA

S'ha tancat les inscripcions en arribar al nombre màxim de places

PROGRAMA DEL CURS

All lectures will focus on how to apply the concepts that are presented using a modern statistical software. All models will be introduced and presented with examples and case studies.

Lecture 1: Data Visualisation

In this introductory lecture, we will use numerous data sets to demonstrate visualisation techniques using the statistical software JMP. JMP is a user-friendly, highly interactive and graphical statistics program and therefore ideally suited for graphical data representation. Among others, we will include heat maps, bubble plots, choropleth maps, 3D scatter plots and volcano plots.

Lecture 2: Linear Models

Analysis of variance (ANOVA) and linear regression are two fundamental statistical models used in numerous applications. Every scientist, engineer or economist who works with data will most likely encounter these linear models in his area of expertise. This lecture will explain the concepts and demonstrate how to apply these linear models to given data sets. Finally, we will use another linear model, ANCOVA, for a case study application.

Lecture 3: Advanced Modelling Techniques

In this lecture, we will use a general linear model as starting point to introduce alternative models. On one hand, we will loosen assumptions to be able to reach more flexible models, the generalized linear models. On the other hand, we will have a look at nonlinear regression models and how to apply them. Finally, we will introduce clustering as an example of an unsupervised modelling technique and apply it in combination with a nonlinear pharmacokinetic model.

Lecture 4: Predictive Modelling

Our focus in this lecture will be on classification methods. We will introduce several classifiers and give examples of their use. For predictive models, model validation is a key concept that we will cover. Typically, one builds several models for predictive purposes, so that one needs to compare their performance. Model comparison measures will be introduced and used for a partition, logistic regression and discriminant analysis.

Lecture 5: Design of Experiments

Design of experiments or DOE is a key tool for product and process improvement and innovation, and for exploring how multiple factors simultaneously influence the performance of a process or the characteristics of a product. We will start with classic designs, like factorial and fractional factorial designs, central composite designs and demonstrate their use in case studies. Then we will cover a more modern approach, named optimal design of experiments, and show its advantages for more involved and complex design problems.