Multiple Hypothesentests

Lecturer: Markus Pauly
Exercises taught by:
Dennis Dobler


General Information

LanguageEnglish
Lectures2 h

Exercises

Office Hours

1h

by appointment


Time and Venue

Lectures Wednesday, 08:00-10:00, HeHo 18/220

Exercises Wednesday, 12:00-14:00, HeHo 18/E60


Exam:

Oral Exams on July 13 as well as October 4, 2016.


General Informations:

Prerequisites:

Suitable for bachelors students in "Mathematische Biometrie", "Mathematik" or "Wirtschaftsmathematik" (each in their last studying year).
Attending students must have passed the lectures on "Elementare Wahrscheinlichkeitsrechnung und Statistik" as well as "Stochastik 1".

Exam:

 

If 60 per cent of all exercise points are achieved and if the participant at least once presented her/his solution of an exercise, then the exam degree will be upgraded to the next best degree (e.g., 2.0 --> 1.7 or 3.7 --> 3.3 etc).


Contents: In applied practice, it is often not sufficient to test only one null hypothesis. For instance, after rejection of ``two groups possess the same location parameter''
the statistician may be interested in the specific ordering of both. Another example is motivated by a 1-factorial analysis of variance (ANOVA). Here, a rejection of the null ``all groups possess the same expected value'' gives rise to testing hypotheses about more specific differences among those. Since testing several null hypotheses with the same data-sets usually leads to cumulated error probabilities,
a theory for the construction of adequate multiple hypothesis tests is mandatory. During this course, such a theory will be developed and motivated on the basis of practical examples. As quality criteria of particular interest the concepts of classical FWER as well as of modern FDR control will be covered.


Exercise Sheets & other Files: Will be available on Moodle. Please apply there for this course.


Literature:

Click here

Semesterapparat:

Click here


Note