Introduction to Generalised Linear Models
Dr Brian Fogarty
Lecturer in Quantitative Social Science
This short course serves as an introduction to Generalised Linear Models in the social sciences. Using R, we will examine the background of Maximum Likelihood Estimation, binary, ordered, unordered outcome, and count data regression models. The emphasis will be on the application of these regression models to social science data. Some basic knowledge of R is required (e.g. how to read in a data file and use basic commands).
Lab A (room 912) | Adam Smith Building |
Wednesday, 13 June 2018 from 10:00 to 13:00
Please register for this training using Eventbrite
Taking Time Seriously: Introduction to Time Series Analysis
Dr Niccole Pamphilis
Lecturer in Quantitative Social Science
This course will introduce participants to the logic and application of time-series analysis. Participants will look at situations where time needs to be accounted for in their models; how to determine the appropriate structure of time (i.e., shocks versus memory); and introduce how to control for time in their models. Statistical software covered will include R and Stata.
Lab M (ASB1105) | Adam Smith Building |
Wednesday, 20 June 2018 from 10:00 to 13:00
Please register for this training using Eventbrite
Introduction to Dimension Reduction through Principal Component and Factor Analysis
Dr Nema Dean
Q-Step Lecturer in Maths & Stats
Both factor analysis and principal component analysis are important tools for reducing the dimensions of datasets with a large number of variables. They are often used to simplify the data structure to allow more complex models to be used later but they can also provide information about hidden structures in the original data. In this course we will look at applying these methods in R and discussing and interpreting the results. Students are welcome to bring their own data sets (in cleaned-up format) to work on. Some working knowledge of R is required prior to the course (e.g. how to read in a data file and use basic commands).
Lab M (ASB1105) | Adam Smith Building |
Wednesday, 27 June 2018 from 10:00 to 13:00
Please register for this training using Eventbrite
Maximum Likelihood Estimation
Dr Niccole Pamphilis
Deputy Director of Q-Step, Lecturer in Quantitative Social Sciences
Email: niccole.pamphilis@glasgow.ac.uk
August 20th-22nd 2018, The University of Glasgow.
Sessions will take place in Lab L, 1113, Adam Smith Building
Please register for this training using Eventbrite
Course Description
This course offers participants an introduction to a range of statistical models beyond those available using standard linear regression analysis. The course will teach students about logit and probit models for dependent variables with binary outcomes as well as ordered models and multinomial models for categorical dependent variables. Participants will learn about the assumptions of each model and how to properly test if the model assumptions hold in addition to expressing model results graphically. Background knowledge of OLS Regression and basic training and understanding of SPSS, R, or Stata are expected.
Monday:
Morning: Review of Regression Analysis
Afternoon: Introduction to Maximum Likelihood Estimation Approach
Tuesday:
Morning: Logit and Probit Models
• Estimation Approach
• Differences Between Models
• Interpretation of Results
• Graphing Results
• Model Diagnostics
Afternoon: Computer Lab Session
• Running Models
• Running Tests
• Graphing: Marginal Effects Plots
Wednesday:
Morning: Ordered and Multinomial Models
• Ordered Models
o Approach
o Interpretation
o Tests
• Multinomial Models
o Approach
o Interpretation
o Tests
Afternoon: Computer Lab Session
• Running Models and Tests
Multi-level Modeling
Professor Mark Tranmer
Professor of Quantitative Social Science
Email: mark.tranmer@glasgow.ac.uk
August 20th-22nd 2018, The University of Glasgow.
Sessions will take place in Lab M, 1105, Adam Smith Building
Please register for this training using Eventbrite.
Course Description:
We will begin by examining the basics of a two-level model. From there we will proceed
to the analysis of other hierarchical structures with several levels and then onto complex
non-hierarchical population structures, including cross-classified and multiple
membership models. We look at examples for the use of multilevel models with
longitudinal data and social network data.
Throughout the course, we will look at examples based on substantive research
questions, and will include hands-on computing sessions to practice the application of the
various approaches. The main software used is R – and also MLwiN with R, through the
packages MLwiN, as will be explained. We will look other R packages that enable more
complex multilevel models to be fitted, including R2WinBUGS and R2OpenBUGS – which
work with WinBUGS and OpenBUGS respectively, and also the MCMCglmm package.
Prerequisites: Students should be familiar with basic statistical methods including OLS
regression analysis. Familiarity with the basics of logistic regression would also be an
advantage, though I will briefly review this, prior to giving details of the multilevel logistic
regression model.
Suggested Texts:
Snijders T and Bosker B (2012) Multilevel Analysis: An Introduction to Basic and
Advanced Multilevel Modelling. Sage Publications.
Further Reading:
Chapter 6 of Crossley N, Bellotti E, Edwards G, Everett MG, Koskinen J, Tranmer M
(2015) Social Network Analysis for Ego-Nets. Sage Publications.
ICPSR Summer Program 2018
Course Topics:
1. Introduction to Multilevel Models.
2. Random Intercept Models.
3. Random Slope Models.
4. Logistic Multilevel Models
5. Other non-linear multilevel models.
6. Cross-Classified Models.
7. Multiple Membership Models.
8. Multilevel Models for Longitudinal Data
9. Use of Multilevel Models with Social Network Data (outline)
10. Alternative Software.
Instructor - Biographical Sketch
Mark Tranmer is Professor of Quantitative Social Science in the School of Social &
Political Sciences University of Glasgow. His methodological research focus began in
multilevel modelling to assess geographical and organisational variations in individual
responses of interest. Recently, he has been interested in developing multilevel
approaches to estimate social network variations in individual level responses. He has
further extended these approaches to assess changes over time in these variations.
These methodological developments highlight his general interest in understanding social
population structure from a geographical, organisational, network, and temporal
perspective. Substantive applications of these methods include assessing individual and
area variations in the take-up of post-compulsory education, inequalities in health and
well-being, variations in crime by local area, variations in civic and political engagement
in the UK and Europe, changes in occupational tie structures in the UK over time, network
variations in hospital waiting times and patient safety, and area, school, and network
variations in drinking, smoking and drug use in six European countries.
He has published in a range of substantive and methodological journals. He was recently
guest co-editor of a special edition of “Social Networks” on Multilevel Social Networks. He
has honorary senior and professorial positions at the Universities of Stirling, Wollongong
(Australia) and USI (Switzerland). He has also taught a range of statistical methods
courses at various levels in the UK and internationally to academic and non-academic
audiences, He leads the Glasgow Quantitative Methods Group (GQMG), which promotes
the interdisciplinary use of quantitative methods in research & teaching.
https://twitter.com/G_Q_M_G
Example publication (involving more advanced multilevel models): Tranmer M., Steel, D.
and Browne, WJ (2014) Multiple-membership Multiple-Classification Models for Social
Network and Group Dependencies. Journal of the Royal Statistical Society: Series A
(Statistics in Society), vol. 177,no. 2, pages 439-455.
EXCEL spreadsheet basics
Professor John McColl
Professor of Learning and Teaching in Statistics
A spreadsheet package, such as EXCEL, provides an excellent platform for recording data, editing and organising them, and running basic queries. It is also possible to carry out many standard statistical analyses within EXCEL. When a more powerful statistical system is required, data may be exported from EXCEL in various formats that are compatible with specialist software packages such as R. This short workshop will provide a hands-on introduction to data entry, manipulation and plotting in EXCEL.
Topics to be covered include:
• Entering and exporting data
• Working with different types of data (numbers, text, dates)
• Conditional formatting
• Formulas and functions
• Absolute and relative addressing
• Summarising data appropriately (mean or median, standard deviation, etc.)
• Plotting data appropriately and editing plots
• Pivot tables
This training will run again in the academic session 2018/2019
Using EXCEL for data management and analysis
Professor John McColl
Professor of Learning and Teaching in Statistics
A spreadsheet package, such as EXCEL, provides an excellent platform for recording data, editing and organising them, and running basic queries. It is also possible to carry out many standard statistical analyses within EXCEL. When a more powerful statistical system is required, data may be exported from EXCEL in various formats that are compatible with specialist software packages such as R. This short workshop will provide a hands-on introduction to data management and analysis in EXCEL. Participants will be assumed to have experience of EXCEL at least equivalent to the skills covered in the EXCEL Spreadsheet Basics workshop.
Topics to be covered include:
• Sorting and filtering data
• Creating and formatting tables
• Databases and database functions
• Combining datasets
• Univariate tests and confidence intervals
• The chi-squared test
• Scatter plots, correlation and regression
This training will run again in the academic session 2018/2019
What is R?
Professor Adrian Bowman
Head of School/Professor of Statistics
R is an open source statistical computing system which has facilities for a very wide range of statistical methods but which is also a very flexible programming environment. There is now a very large user community and a considerable collection of additional packages is available for specialist topics. This short course aims to provide a broad introduction to the system. The topics covered will include examples of standard analyses, including linear (and logistic regression) and multivariate data. There will be a strong emphasis on graphics. The course will be based around a number of case studies and there will be an opportunity for practical use of the system. The course will not involve a detailed exposition of syntax, as prepared scripts will be available as a starting point for examples and practical work. The broad philosophy of R and the very wide range of facilities offered by the environment will be outlined. The course is primarily aimed at staff and postgraduate students in the social sciences.
This training will run again in the academic session 2018/2019.
What is R Again?
Professor Adrian Bowman
Head of School/Professor of Statistics
The aim of this short course is to explore some of the more advanced aspects of R as a environment for the exploration and analysis of data. Participants will be invited to express interest in particular topics and this will have an influence on those chosen as the focus of the examples. Possibilities include random effect models, flexible regression models, spatial analysis and the use of R as a programming environment. There may be an opportunity for participants to bring their own data, depending on the size of the group. The target audiences for the course are research students and staff in the social sciences who have an introductory-level knowledge of R. This includes those who have attended the 'What is R?' session.
This training will run again in the academic session 2018/2019
Introduction to Cluster Analysis
Dr Nema Dean
Q-Step Lecturer in Maths and Stats
In this course we will explain the ideas behind some classical clustering techniques and demonstrate them in R on some interesting urban and crimonological data sets. The course will try to give you an intuition about the methods to enable easier decision making when using them and interpreting their results. Some clustering visualisation methods will also be discussed. Students are welcome to bring their own data sets (in cleaned-up format) to work on. Some working knowledge of R is required prior to the course (e.g. how to read in a data file and use basic commands).
This course will run again in the academic session 20018/2019