-
- Mod 4 - Log Reg
- 4.1 Overview
- 4.2 Odds, Odds Ratios and Exponents
- Quiz A
- 4.3 A General Model
- 4.4 Log Reg Model
- 4.5 Logistic Equations
- 4.6 How good is the model?
- 4.7 Multiple Explanatory Variables
- 4.8 Methods of Log Reg
- 4.9 Assumptions
- 4.10 Example from LSYPE
- 4.11 Log Reg on SPSS
- 4.12 SPSS Log Reg Output
- 4.13 Interaction Effects
- 4.14 Model Diagnostics
- 4.15 Reporting the Results
- Quiz B
- Exercise
-
- Mod 4 - Log Reg
Using Statistical Regression Methods in Education Research
4.1 Overview
|
In the last two modules we have been concerned with analysis where the outcome variable (sometimes called the dependent variable) is measured on a continuous scale. However many of the variables we meet in education and social science more generally have just a few, maybe only two categories. Frequently we have only a dichotomous or binary outcome. For example this might be whether a student plans to continue in full-time education after age 16 or not, whether they have identified Special Educational Needs (SEN), whether they achieve a certain threshold of educational achievement, whether they do or do not go to university, etc. Note that here the two outcomes are mutually exclusive and one must occur. We usually code such outcomes as 0 if the event does not occur (e.g. the student does not plan to continue in FTE after the age of 16) and 1 if the event does occur (e.g. the student does plan to continue in FTE after age 16).
This module first covers some basic descriptive methods for the analysis of binary outcomes. This introduces some key concepts about percentages, proportions, probabilities, odds and odds-ratios. We then show how variation in a binary response can be modelled using regression methods to link the outcome to explanatory variables. In analysing such binary outcomes we are concerned with modelling the probability of the event occurring given the level/s of one or more explanatory (independent/predictor) variables. This module is quite difficult because there are many new concepts in it. However if you persevere through the rationale below, you will find (hopefully!) that the examples make sense of it all. Also, like all things, the concepts and applications will grow familiar with use, so work through the examples and take the quizzes.
As with previous modules you can follow us through the real-world examples that we use. You may now be familiar with the main adapted version of the LSYPE dataset but we also have a more specialised one for use with this module – the one we used in the previous module. We recommend that you retrieve them from the ESDS website – playing around with them will really help you to understand this stuff!
|
