Postgraduate Course: Multi-Level Modelling in Social Science (PGSP11424)
Course Outline
| School | School of Social and Political Science |
College | College of Arts, Humanities and Social Sciences |
| Credit level (Normal year taken) | SCQF Level 11 (Postgraduate) |
Availability | Available to all students |
| SCQF Credits | 20 |
ECTS Credits | 10 |
| Summary | Multilevel models are becoming an increasingly popular method of analysis in many areas of social science, medicine and natural science, and there are many situations where an improved analysis is obtained compared to conventional methods such as ANOVA or multiple regression. Potential advantages include:
- the scope for wider inference: for example in a study of school attainment, results can be related to a population of schools rather than just those assessed;
- more appropriate mean estimates, when the effect of spurious outlying results for small groups are reduced;
- a more efficient analysis with smaller standard errors, particularly when there are few observations per group;
- avoidance of problems caused by missing outcomes: this is an advantage in longitudinal studies (for example panel studies) where there are often dropouts;
- use of more appropriate variances and correlations: for example in a longitudinal analysis the correlation between observations on the same person may become less for measurements that are further apart in time.
The course enables students to understand and use multilevel models mainly in the context of social science, but examples are also given from medicine and some aspects of biological science. The focus is on multilevel models for quantitative, binary and multinomial outcomes, with shorter sessions on models for ordinal and count outcomes. The importance of multilevel modelling for longitudinal data is explained. Analysis is illustrated using the package MLwiN (dedicated to multilevel modelling and available free to academics and university students). Lectures are combined with practical sessions in order to reinforce concepts. |
| Course description |
Multilevel models are becoming an increasingly popular method of analysis in many areas of social science, medicine and natural science, and there are many situations where an improved analysis is obtained compared to conventional methods such as analysis of variance or multiple regression. Potential advantages include:
- the scope for wider inference: for example in a study of school attainment, results can be related to a population of schools rather than just those assessed;
- more appropriate mean estimates, when the effect of spurious outlying results for small groups are reduced;
- a more efficient analysis with smaller standard errors, particularly when there are few observations per group;
- avoidance of problems caused by missing outcomes: this is an advantage in longitudinal studies (for example panel studies) where there are often dropouts;
- use of more appropriate variances and correlations: for example in a longitudinal analysis the correlation between observations on the same person may become less for measurements that are further apart in time.
The course enables you to understand and use multilevel models mainly in the context of social science, but examples are also given from medicine and some aspects of biological science. The focus is on multilevel models for quantitative, binary and multinomial outcomes, with further sessions on models for ordinal and count outcomes. The importance of multilevel modelling for longitudinal data is explained. Analysis is illustrated using the statistical package Stata.
Most of your learning will be in practical work. At each of the weekly practical sessions, two teachers will be available for consultation. There are also fortnightly practical sessions devoted to the presentation of the results of multilevel models.
|
Information for Visiting Students
| Pre-requisites | None |
| High Demand Course? |
Yes |
Course Delivery Information
|
| Academic year 2022/23, Available to all students (SV1)
|
Quota: 40 |
| Course Start |
Semester 2 |
Timetable |
Timetable |
| Learning and Teaching activities (Further Info) |
Total Hours:
200
(
Seminar/Tutorial Hours 20,
Programme Level Learning and Teaching Hours 4,
Directed Learning and Independent Learning Hours
176 )
|
| Assessment (Further Info) |
Written Exam
0 %,
Coursework
100 %,
Practical Exam
0 %
|
| Additional Information (Assessment) |
There are three components to the assessment:
1) A formative assessment early in the course using single level regression. This allows students a chance to check their familiarity with relevant statistical software, and to get early feedback on the presentation of statistical results. Students will be offered online tuition in R prior to the course.
2) The mid-term coursework will involve students fitting, reporting and briefly interpreting a set of multilevel models which will use a continuous outcome. There will be a series of short questions to guide you through the type of analysis you need to complete. Models covered will include variance component, random intercept and random slope. Sections of the exercise will also be concerned with providing brief substantive interpretations of the results you present
3) An end-of-course practical project, worth 70% of the total marks for the course. Students are given access to a general social science dataset (such as the European Social Survey) which is suitable for multilevel analysis. Students are then asked to identify an appropriate dependent variable and develop a research question involving at least 5 independent variables. Alternatively, students may find their own data and research question. They are required to write a report of no more than 3250 words which answers their research question and uses at least one variance component model, one random intercept model and one random slope model. |
| Feedback |
Feedback is provided within 15 working days. |
| No Exam Information |
Learning Outcomes
On completion of this course, the student will be able to:
- Critically evaluate when the use of multilevel models is appropriate in social science research.
- Exercise autonomy in developing research questions than can be addressed using multilevel models.
- Estimate theoretically informed multilevel models in an appropriate software package .
- Interpret and communicate the results of multilevel models to a variety of audiences (including non-technical practitioners) with understanding of their limitations and assumptions.
|
Reading List
Dale, A. and Davies, R. (1994), Analyzing Social and Political Change, London: Sage.
Goldstein, H (2011) Multilevel Statistical Models (fourth edition), London: Edward Arnold.
Quintelier, E. (2010), 'The effect of schools on political participation: a multilevel logistic analysis', Research Papers in Education, 25, pp. 137-154.
Snijders, T and Bosker, R (1999), Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modelling, London: Sage Publishing.
Steenbergen, M. and Jones, B. (2002), 'Modeling multilevel data structures', American Journal of Political Science,46, pp. 218-237.
Sun, L., Bradley, K. D. and Akers, K. (2012), 'A multilevel modelling approach to investigating factors impacting science achievement for secondary school students: PISA Hong Kong sample', International Journal of Science Education, 34, pp. 2107-2125.
Whitworth, A. (2012), 'Inequality and crime across England: a multilevel modelling approach', Social Policy and Society, 11, pp 27-40.
|
Additional Information
| Graduate Attributes and Skills |
Advanced understanding of and competence with multilevel modelling.
Capacity to apply multilevel modelling to social scientific questions. |
| Keywords | Not entered |
Contacts
| Course organiser | Dr Orian Brook
Tel:
Email: |
Course secretary | Mrs Beth Richardson-Mills
Tel: (0131 6)51 1659
Email: |
|
|
University Homepage