Skip to main content
Fig. 5 | BMC Medical Research Methodology

Fig. 5

From: Minimizing bias in massive multi-arm observational studies with BCAUS: balancing covariates automatically using supervision

Fig. 5

Comparison between BCAUS and BART ATE estimates on the diabetes dataset. Since ground truth for the diabetes dataset is unknown, we compare BCAUS estimates versus BART estimates. ATE estimated using BCAUS-IPTW plotted against ATE estimates from BART (green dots) for 133 intervention arms. Units are percentage changes in glycated hemoglobin i.e. HbA1c. Straight line represents y = x. Mean absolute error between the two estimates is 0.04

Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us