2011/12 – Binární logistická regrese a příbuzné metody

Název kurzu: Binary Logistic Regression and Related Methods

Modul: kvantitativní

Kód: PSY028e a SOC931e

Vyučující: Prof. Donald J. Treiman

Termín konání: 10. 14. dubna  2012

úterý 10.4., 9-12, 13-16, PC 26

středa 11.4.,9-12, 13-16, PC 26

čtvrtek 12.4., 12-15, 16-19,PC 26

pátek 13.4., 9-12, 13-16, PC 26

sobota 14.4., 9-12,13-16, PC 26

Vyučovací jazyk: angličtina



This is a course in how to do theoretically informed quantitative social research, using a body of statistical procedures used to study limited dependent variables. We will cover

Binomial logistic regression, which is used for the analysis of dichotomous variables such as whether or not one votes.
Discrete-time event history analysis of binary outcomes—for example, the age at which people marry.
Multinomial logistic regression, which is used for the analysis of polytomous (categorical) dependent variables such as the type of secondary school people enter (e. g., academic, vocational, or none).
Tobit analysis, used to study “censored” dependent variables, that is, variables whose theoretical range extends beyond the measured range—for example, the propensity to have extramarital affairs (among those who have never had an extramarital affair, for some people it might be out of the question while for others it is just that an opportunity hasn’t presented itself).
Scale construction—the creation of variables to measure traits such as happiness, political conservatism, etc.
Ordinal logistic regression, which is used to analyze “ordinal” independent variables, where we can order the response categories but don’t know how far apart successive categories are from one another (for example, “Are you very happy, somewhat happy, or not too happy?”).

The course is designed for those who already have had a statistics course that covered ordinary least squares multiple regression.  Although the focus is on the interpretation of statistical material, it necessarily presents the statistical procedures along the way.  But this is not a statistics course; the emphasis will be on using these procedures to draw substantive conclusions about how the social world works. If you already have had some exposure to logistic regression, so much the better.  If you have not, but you understand multiple regression, you probably will be comfortable since the techniques covered are conceptually (although not operationally) similar to multiple regression.