Yuling Yao

Yuling Yao

Auto-generated profile, based on publicly available information

Yuling Yao

Flatiron Institute, New York, USA

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 September 2018 in Bayesian Analysis
Verified
Journal Article
Open access
PDF file
YYYuling Yao
Yuling Yao
AVAki Vehtari
Aki Vehtari
DSDaniel Simpson
Daniel Simpson
AGAndrew Gelman
Andrew Gelman
Bayesian model averaging is flawed in the M-open setting in which the true data-generating process is not one of the candidate models being fit. We take the idea of stacking from the point estimation literature and generalize to the combination of predictive distributions. We extend the utility function to any proper scoring rule and use Pareto smoothed importance sampling to efficiently compute the required leave-one-out posterior distributions. We compare stacking of predictive distributions to several alternatives: stacking of means, Bayesian model averaging (BMA), Pseudo-BMA, and a variant of Pseudo-BMA that is stabilized using the Bayesian bootstrap. Based on simulations and real-data applications, we recommend stacking of predictive distributions, with bootstrapped-Pseudo-BMA as an approximate alternative when computation cost is an issue.

1 December 2022 in Bayesian Analysis
Verified
Journal Article
Open access
PDF file
YYYuling Yao
Yuling Yao
GPGregor Pirš
Gregor Pirš
AVAki Vehtari
Aki Vehtari
AGAndrew Gelman
Andrew Gelman
Stacking is a widely used model averaging technique that asymptotically yields optimal predictions among linear averages. We show that stacking is most effective when model predictive performance is heterogeneous in inputs, and we can further improve the stacked mixture with a hierarchical model. We generalize stacking to Bayesian hierarchical stacking. The model weights are varying as a function of data, partially-pooled, and inferred using Bayesian inference. We further incorporate discrete and continuous inputs, other structured priors, and time series and longitudinal data. To verify the performance gain of the proposed method, we derive theory bounds, and demonstrate on several applied problems.

10 December 2020 in Journal of Physics G: Nuclear and Particle Physics
Verified
Journal Article
Closed access
AGAndrew Gelman
Andrew Gelman
YYYuling Yao
Yuling Yao
Every philosophy has holes, and it is the responsibility of proponents of a philosophy to point out these problems. Here are a few holes in Bayesian data analysis: (1) the usual rules of conditional probability fail in the quantum realm, (2) flat or weak priors lead to terrible inferences about things we care about, (3) subjective priors are incoherent, (4) Bayesian decision picks the wrong model, (5) Bayes factors fail in the presence of flat or weak priors, (6) for Cantorian reasons we need to check our models, but this destroys the coherence of Bayesian inference. Some of the problems of Bayesian statistics arise from people trying to do things they should not be trying to do, but other holes are not so easily patched. In particular, it may be a good idea to avoid flat, weak, or conventional priors, but such advice, if followed, would go against the vast majority of Bayesian practice and requires us to confront the fundamental incoherence of Bayesian inference. This does not mean that we think Bayesian inference is a bad idea, but it does mean that there is a tension between Bayesian logic and Bayesian workflow which we believe can only be resolved by considering Bayesian logic as a tool, a way of revealing inevitable misfits and incoherences in our model assumptions, rather than as an end in itself.

21 August 2020 in GeoHealth
Verified
Journal Article
Open access
PDF file
AGAlexander van Geen
Alexander van Geen
YYYuling Yao
Yuling Yao
TETyler Ellis
Tyler Ellis
AGAndrew Gelman
Andrew Gelman
The roof and spire of Notre‐Dame cathedral in Paris that caught fire and collapsed on 15 April 2019 were covered with 460 t of lead (Pb). Government reports documented Pb deposition immediately downwind of the cathedral and a twentyfold increase in airborne Pb concentrations at a distance of 50 km in the aftermath. For this study, we collected 100 samples of surface soil from tree pits, parks, and other sites in all directions within 1 km of the cathedral. Concentrations of Pb measured by X‐ray fluorescence range from 30 to 9,000 mg/kg across the area, with a higher proportion of elevated concentrations to the northwest of the cathedral, in the direction of the wind prevailing during the fire. By integrating these observations with a Gaussian process regression model, we estimate that the average concentration of Pb in surface soil downwind of the cathedral is 430 (95% interval, 300–590) mg/kg, nearly double the average Pb concentration in the other directions of 240 (95% interval, 170–320) mg/kg. The difference corresponds to an integrated excess Pb inventory within a 1 km radius of 1.0 (95% interval, 0.5–1.5) t, about 0.2% of all the Pb covering the roof and spire. This is over 6 times the estimated amount of Pb deposited downwind 1–50 km from the cathedral. To what extent the concentrated fallout within 1 km documented here temporarily exposed the downwind population to Pb is difficult to confirm independently because too few soil, dust, and blood samples were collected immediately after the fire.

1 January 2022 in International Review of Economics & Finance
Verified
Journal Article
Closed access
YSYUSUNG SU
YUSUNG SU
DLDonald Lien
Donald Lien
YYYuling Yao
Yuling Yao

17 May 2021 in arXiv
Verified
Preprint
Open access
PDF file
YYYuling Yao
Yuling Yao
A general challenge in statistics is prediction in the presence of multiple candidate models or learning algorithms. Model aggregation tries to combine all predictive distributions from individual models, which is more stable and flexible than single model selection. In this article we describe when and how to aggregate models under the lens of Bayesian decision theory. Among two widely used methods, Bayesian model averaging (BMA) and Bayesian stacking, we compare their predictive performance, and review their theoretical optimality, probabilistic interpretation, practical implementation, and extensions in complex models.