Two countries, lots of models and one secret weapon: Two applications of the "many models" workflow
- Two countries, lots of models and one secret weapon: Two applications of the "many models" workflow
- Speaker: Ben Matthews # University of Edinburgh
- Hosted by
- Introduced by
- Date and Time
- 25th Mar 2019 13:00 - 25th Mar 2019 14:15
- Practice Suite 1.12, Chrystal Macmillan Building
Crime has fallen across the English-speaking world from the early 1990s to the early 2010s. Criminologists are still struggling to explain this decline, and recently some have suggested the adoption of comparisons of convictions trends across cohorts and countries as a vital but missing piece in the descriptive evidence base needed to understand the international crime drop.
In his talk, Ben will demonstrate two applications of the 'many models' workflow. It offers a way for data analysts to better understand complex datasets by fitting multiple simple models to subsets of data. Ben will apply this workflow, first, to the problem of understanding how the recent crime drop is reflected in individual patterns of criminal careers by fitting Latent Class Growth Models (LCGMs) to convictions data from multiple birth cohorts; and second, to explore variation in convictions trajectories produced by LCGMs by fitting models to multiple bootstrap draws from an original dataset. Ben will focus both on the benefits and limitations of the many models approach in these two applications, as well as how this workflow can be implemented using R, Mplus and the MplusAutomation R package to fit and analyse batches of LCGMs using functional programming.
This work is part of the Understanding Inequalities project which aims to explore the causes and consequences of inequalities in crime, education, housing and the labour market Scottish society and beyond.
The seminar will take place on 25th March from 1pm in the Practice Suite (1.12) on the first floor of the Chrystal Macmillan Building.