SA Branch webinar: Propensity score techniques in multiple treatment framework: the estimation of neighbourhood effect

Branch Meeting - Thursday, 3rd September 2020

The South Australian Branch of the Statistical Society would like to invite you to one of two September meetings of the 2020 program.

Virtual Venue via Zoom meeting:

https://unisa.zoom.us/j/99936373140?pwd=T2dPK0tJZFJaeUhTYzhjV2tuencxdz09

Password: 107151

or Join from dial-in phone line:

Dial: +61 2 8015 2088

Meeting ID: 999 3637 3140

Speaker: Dr Margherita Silan, Padua University

Topic: Propensity score techniques in multiple treatment framework: the estimation of neighbourhood effect

Abstract

The study of neighbourhood effects on health conditions has gained attention exponentially during last decades. This topic is even more important from a public health perspective when the focus is on health conditions of elderly, whom spend more time in their neighbourhood than young people. The study estimates neighbourhood effects on elderly health outcomes in Turin (a city located in north Italy) using data coming from the Turin Longitudinal Study, that contains linked individual data from censuses and administrative health data flows. Individuals are not randomly distributed among neighbourhoods, the residential area is chosen also on the basis of individual socio-economic characteristics. This causes a selection bias, that may confound comparisons between the distribution of health outcomes among neighbourhoods. In order to balance observable confounders and to make different populations living in different neighbourhoods comparable, we adopted a propensity score approach. The original methodological contribution of our work is the adaptation of propensity score techniques to a framework with many treatments (that are neighbourhoods in this application). Indeed, Turin may be divided according to three geographical partitions in 10, 23 or 94 neighbourhoods. We proposed a novel method that consists on a Matching on Poset based Average Rank for Multiple Treatments (MARMoT), which has revealed to be really useful to improve the covariates’ balance between groups and pretty fast with respect to other approaches, that have revealed to be unpractical in applications with many treatments.

Biography

Margherita Silan is a postdoctoral research fellow at the Department of Statistical Sciences, Padua University. She received a PhD degree from the Padua University, in 2019. Margherita is also the winner of the Italian Statistical Society annual award for best PhD in Applied Statistics. Her research interests include causal inference in multiple treatment frameworks, composite indicators and partially ordered set theory. The application areas are public health, ageing and gender equality.

Feel free to forward this meeting notice to colleagues, all welcome.