Website
Module Guide before WS 2016/17

Modul MA4962-KP05

Generalized Linear Models (VLModKP05)

Duration:


1 Semester
Turnus of offer:


irregularly
Credit points:


5
Course of studies, specific field and terms:
  • Bachelor CLS 2016 (optional subject), mathematics, 5th or 6th semester
  • Master CLS 2016 (optional subject), mathematics, 1st, 2nd, or 3rd semester
Classes and lectures:
  • Generalized Linear Models (exercise, 1 SWS)
  • Generalized Linear Models (lecture, 2 SWS)
Workload:
  • 35 Hours in-classroom work
  • 15 Hours exam preparation
  • 45 Hours programming
  • 25 Hours private studies
  • 30 Hours work on project
Contents of teaching:
  • General overview of generalized linear models (GLM): - link and response function, - GLM algorithms: Newton-Raphson, Fisher Scoring, iterated weighted least squares, - convergence, - quality of the adaption, - residuals
  • Continuous response models: Gaussian, log-normal, Gamma, log-Gamma for survival analysis, inverse Gaussian
  • Dichotomous response models: logit, probit, cloglog
  • Count data: Poisson, negative binomial with over- and underdispersion
  • Ordinal response models: proportional odds model
  • Disordered categorial response models: Multinomial logit and probit model
  • Censored continuous response models: Tobit model
Qualification-goals/Competencies:
  • The students are able to explain the theoretical bases of generalized linear models (GLM).
  • They are able to explain areas of application for GLM.
  • They are able to select a suitable GLM.
  • They are able to estimate GLMs in R.
  • They are able to explain the R source code in a presentation.
  • They are able to judge the results of GLMs in R critically.
  • They are able to evaluate algorithmic challenges of GLMs.
  • They are able to explain conceptual problems of GLMs for categorial response variables.
  • They are able to implement GLM in R.
  • They are able to apply regression diagnostics to GLMs and to judge the results.
  • They are able to describe the most important estimation algorithms for GLMs.
  • They are able to list the statistical properties of GLMs.
Grading through:
  • Viva Voce or test
Requires:
Responsible for this module:
  • Prof. Dr. med. Peter König
Teachers:
  • Prof. Dr. rer. biol. hum. Inke König
Literature:
  • Agresti, Alan: Foundations of Linear and Generalized Linear Models - Wiley, 2015
Language:
  • English, except in case of only German-speaking participants
Notes:

Prüfungsvorleistungen können zu Beginn des Semesters festgelegt werden. Sind Vor- leistungen definiert, müssen diese vor der Erstprüfung erbracht und positiv bewertet worden sein.

Letzte Änderung:
27.9.2021