Universität Wien FIND
Warning! The directory is not yet complete and will be amended until the beginning of the term.

400016 SE Seminar for Doctoral Candidates: Methods (2011S)

Applied Logistic Regression and other advanced regression models

Continuous assessment of course work

Tuesday, 08.03.2011-12.04.2011, 10:00 -14:00
Place: NIG, Kursraum B, EG

Registration (Anmeldung): please e-mail markus.wagner@univie.ac.at to register for the course.

This course is open to MA and PhD students!

Details

Language: English

Lecturers

Classes

Currently no class schedule is known.

Information

Aims, contents and method of the course

Almost all introductions to quantitative methods for the social sciences end with linear regression models. However, these generally work best with continuous dependent variables, for example income, GDP or IQ scores. Most real-world social science questions have simpler dependent variables, so ones that have just a limited number of possible outcomes. Turnout and vote choice are well-known examples from political science.

This course provides a practical and applied treatment of regression using such dependent variables. For such variables, different regression techniques need to be used, and this course will introduce you to the use of logit models for binary, nominal and ordinal outcomes.

By the end of this course, you will be able to construct and interpret regression models for categorical dependent variables. You will have a firm understanding of the assumptions and the interpretation of the model. You will also be aware of the differences between these models and OLS regression; you will be able to use various types of independent variables; you will know how to present the results in a scholarly paper; and you will be able to identify and address possible dangers and problems. You will also be able to evaluate critically such regression models used in scholarly journals.

The course will cover all these models in six sessions. We will begin by reviewing basic statistical concepts, in particular OLS regression, before considering the need for other types of regression models. The next two sessions will consider models for binary outcomes and introduce the statistical techniques that can also be used for ordered and nominal outcomes. The next two sessions will cover ordinal and nominal outcomes, which share many features of the models for binary outcomes. A final session will provide some insight into further topics (e.g. interaction effects, count outcomes) depending on the interests of the students.

Each class will be a mixture of short lectures/talks and computer exercises. Class exercises and homework will be carried out using Stata. This programme will be introduced in the first session, and no prior knowledge of it is required. Training in quantitative methods up to very basic OLS regression is required for this course, but detailed knowledge is not expected.

The class will be held in English.

Schedule

8 March, Session 1: Introduction to Stata; review of OLS and why we need logit and probit models
15 March, Session 2: Regression with binary outcomes (1): interpretation
22 March, Session 3: Regression with binary outcomes (2): hypothesis testing and model fit
29 March, Session 4: Regression with ordinal outcomes: ordered logit/probit
5 April, Session 5: Regression with nominal outcomes: multinomial logit
12 April, Session 6: Additional topics and review

All classes will take place in Kursraum B in the Neue Institutsgebäude (NIG).

Assessment and permitted materials

- Four homework and problem sets, to be submitted at four set dates (40%)

- 10-15 page problem-set paper on main concepts and interpretation of results, assigned after the last class OR 10-15 page paper on main concepts, challenges and interpretation of results based on a freely chosen research question (for example linked to a master's or doctoral thesis) (50%)

- Continuous assessment of class participation (10%)

Minimum requirements and assessment criteria

Examination topics

Reading list

J. Scott Long, Regression Models for Categorical and Limited Dependent Variables, Sage: Thousand Oaks, 1997.

J. Scott Long and Jeremy Freese, Regression Models for Categorical and Dependent Variables using Stata, Stata Press: College Station, 2003. (Newer edition also available.)

Scott Eliason, Maximum Likelihood Estimation: Logic and Practice (Quantitative Applications in the Social Sciences), Thousand Hills: Sage, 1993.

John G. Orme and Terri Combs-Orme, Multiple Regression with Discrete Dependent Variables, Oxford: OUP, 2009.

Association in the course directory

Last modified: Fr 31.08.2018 08:58