In a classic analysis from 1897, the French sociologist Émile Durkheim observed that suicide rates appeared to cluster in space. However, suicide was correlated with social integration/fragmentation which also exhibited similar clustering patterns. Gabriel Tarde, a contemporary of Durkheim, argued for additional factors due to imitation and geographic connectivity. How do you handle model properties in space which are correlated? How do you remove/detect effects that are correlated? This talk will briefly discuss these issues using a contemporary dataset describing several aspects of deliberate self-harm in New Zealand. Other stories will include examining redacted paper reviews on Grammatical Evolution and researcher bias when comparing models in Software Effort Estimation.
Last modified: Tuesday, 08-Sep-2015 08:55:52 NZST
This page is maintained by the seminar list administrator.