SSA ACT Branch will be holding its first meeting in 2025 on Tuesday 25th February in person and via Zoom. There will be the opportunity to meet for dinner afterwards in Canberra City.
Please forward to any non-members who may be interested.
Venue: Room 3.03 in Marie Reay Teaching Centre, Australian National University (MRTC ANU), or via Zoom.
Date: Tuesday 25 February 2025
Time: 6:00pm [Note – 15 minutes later than usual time due to room booking restrictions]
Zoom link: https://anu.zoom.us/j/89379372884?pwd=3ljyfGTWk6afZurbkjVE1Kmm3rRmLb.1
Meeting ID: 893 7937 2884
Password: 410204
The zoom link will be open shortly before 6.00pm. RSVP is not required. Full zoom details given at the end of the email.
Speaker: Bronwyn Loong, Research School of Finance, Actuarial Studies and Statistics, College of Business and Economics, Australian National University (joint work with Raymond Liu and Gen Nowak)
Topic: Treatment effect estimation with entropy balancing in observational studies
To establish causal effects in an observational study, it is desirable to replicate the characteristics of a randomised control trial. Entropy balancing (EB) is a data pre-processing technique whereby control group data are reweighted to match the covariate moments in the treatment group. The entropy balancing scheme iteratively searches for the set of weights that satisfy a set of pre-specified balance constraints. Numerous previous studies have evaluated entropy balancing against propensity score based methods for causal inference in observational studies. Yet very few studies have compared entropy balancing with ordinary least squares (OLS) estimation of the treatment effect. In this talk, we review the literature on entropy balancing and present some of our simulation study results to compare EB with OLS. Initial simulation results showed EB had no advantage over OLS when the outcome model is correctly specified. Moreover, the EB algorithm failed to converge in some simulation scenarios. Further simulation studies in the presence of unmeasured confounders showed EB weighted estimates were less biased than OLS estimates in some situations. Further theoretical work is required to identify the exact conditions under which EB will outperform OLS. Given the computational requirements of the EB algorithm where a weighting solution is not guaranteed, this work is important to establish the benefits of EB for causal effect estimation.
Biography: Bronwyn Loong is a Senior Lecturer in Statistics at ANU in the College of Business and Economics. Her research interests are in missing data, data confidentiality, causal inference and Bayesian modelling. Bronwyn enjoys working on applied problems with particular applications in medicine and healthcare.
Dinner: After the talk we will be holding a dinner at 7.30pm at SoLita Pizzeria & Pasta Bar, Baileys Corner, 143 London Circuit, Canberra City (https://solita.com.au/)
If you are interested in attending the dinner, please let us know by 5pm Monday 24 February by entering your details at SSA Canberra Branch dinner attendance sheet or contacting Warren Muller (warren.muller@csiro.au ; 0407 916 868). Please regard this as a firm commitment, not just an intention.
NOTE: We are offering discounts to SSA early career and student members who attend dinner! For this meeting, dinners will be a fixed charge of $10 for student members and $20 for early career members.
@StatSocAus
