130 likes | 136 Views
This session provides an overview of statistical models and frequently used advanced models in regression analysis. Topics covered include multilevel modelling, longitudinal panel data analysis, time-series cross-sectional data analysis, and survival/duration models. The session also offers learning resources such as software packages, online courses, and readings with replication codes.
E N D
Look ahead: Advanced modelling and learning resources Dr. Yang Hu yang.hu@Lancaster.ac.uk Dr. Stuart Bedston s.bedston@lancaster.ac.uk
Overview of the session • Statistical models: A brief overview • Frequently used advanced models • Learning resources
Frequently used ones • Multilevel modelling (data with hierarchical structure) • Longitudinal panel data analysis: Unit-specific changes and continuities over time. • Time-series cross-sectional data: Pseudo-panel – over time change • Survival/duration model: time-dependence of event occurrence
Multilevel modelling • Data: Hierarchical clustering of data (aka. hierarchical regression) • Country county local authority neighborhood family individual • Question: What are the the individual/regional level correlates of fear for crime?
Longitudinal panel models • Data: each individual/unit is observed at multiple time points • Tracing individual-level/unit-level (causal) changes over time (accounting for individual-level heterogeneity) • Often used for the identification of casual relationships Question: Motherhood penalty?
Cross-sectional time-series analysis • Data: Repeated observation of a sampling frame, but not the same individuals/units within the frame. • Example datasets: British Crime Survey, British Social Attitudes, European Social Survey. • Question: Changing crime rate in Britain over time (fixation on the frame, not the specific units).
Survival / Duration Analysis • When your question is about what affects the risk over time of an event occurring • Example: time to missing person being found/case closed • Survival Analysis can handle people going missing and never being found. • Can handle different types of events: dead, alive, stopped • In my work, I have looked at mothers who return to the family courts i.e. they have had one set of children potentially removed from their care and now the local authority wishes to remove a new child.
Learning resources – software packages • R – this page provides a comprehensive list of R learning resources: https://stats.stackexchange.com/questions/138/free-resources-for-learning-r • Stata – The Stata official User Manual is a good place to start: https://www.stata.com/manuals13/u.pdf
Learning resources – Online courses • UCLA – a very comprehensive site with detailed examples: https://stats.idre.ucla.edu/ • UPenn: https://onlinecourses.science.psu.edu/stat414/node/213 • Bristol – Centre for Multilevel Modelling: http://www.bristol.ac.uk/cmm/ • Cathie Marsh Institute, Manchester: https://www.cmist.manchester.ac.uk/ • National Centre for Research Methods: https://www.ncrm.ac.uk/
Learning resources – Readings and example analysis (Quantitative) social science journals without paywall, but with replication codes • Demographic Research: https://www.demographic-research.org/default.htm (with replication codes) • Sociological Science: https://www.sociologicalscience.com/category/articles/ • PlosOne: http://journals.plos.org/plosone/
Learning resources – The example dataset British Crime Survey 2007-2008: Unrestricted Access Teaching Dataset https://discover.ukdataservice.ac.uk/catalogue?sn=6891#administrative