TY - GEN
T1 - Linked causal variational autoencoder for inferring paired spillover effects
AU - Rakesh, Vineeth
AU - Guo, Ruocheng
AU - Moraffah, Raha
AU - Agarwal, Nitin
AU - Liu, Huan
N1 - Funding Information:
This material is based upon work supported by the National Science Foundation (NSF) grant 1614576, and the Office of Naval Research (ONR) grant N00014-17-1-2605.
Funding Information:
of units. Specifically, given a pair of units u and u¯, their individual treatment and outcomes, the encoder network of LCVA samples the confounders by conditioning on the observed covariates of u, the treatments of both u and u¯ and the outcome of u. Using a network of users from job training dataset (LaLonde (1986)) and co-purchase dataset from Amazon e-commerce domain, we show that LCVA is significantly more robust than existing methods in capturing spillover effects. ACKNOWLEDGMENTS This material is based upon work supported by the National Science Foundation (NSF) grant 1614576, and the Office of Naval Research (ONR) grant N00014-17-1-2605. REFERENCES
Publisher Copyright:
© 2018 Association for Computing Machinery.
PY - 2018/10/17
Y1 - 2018/10/17
N2 - Modeling spillover effects from observational data is an important problem in economics, business, and other fields of research. It helps us infer the causality between two seemingly unrelated set of events. For example, if consumer spending in the United States declines, it has spillover effects on economies that depend on the U.S. as their largest export market. In this paper, we aim to infer the causation that results in spillover effects between pairs of entities (or units); we call this effect as paired spillover. To achieve this, we leverage the recent developments in variational inference and deep learning techniques to propose a generative model called Linked Causal Variational Autoencoder (LCVA). Similar to variational au-toencoders (VAE), LCVA incorporates an encoder neural network to learn the latent attributes and a decoder network to reconstruct the inputs. However, unlike VAE, LCVA treats the latent attributes as confounders that are assumed to affect both the treatment and the outcome of units. Specifically, given a pair of units u and u, their individual treatment and outcomes, the encoder network of LCVA samples the confounders by conditioning on the observed covariates of u, the treatments of both u and u and the outcome of u. Once inferred, the latent attributes (or confounders) of u captures the spillover effect of u on u. Using a network of users from job training dataset (LaLonde (1986)) and co-purchase dataset from Amazon e-commerce domain, we show that LCVA is significantly more robust than existing methods in capturing spillover effects.
AB - Modeling spillover effects from observational data is an important problem in economics, business, and other fields of research. It helps us infer the causality between two seemingly unrelated set of events. For example, if consumer spending in the United States declines, it has spillover effects on economies that depend on the U.S. as their largest export market. In this paper, we aim to infer the causation that results in spillover effects between pairs of entities (or units); we call this effect as paired spillover. To achieve this, we leverage the recent developments in variational inference and deep learning techniques to propose a generative model called Linked Causal Variational Autoencoder (LCVA). Similar to variational au-toencoders (VAE), LCVA incorporates an encoder neural network to learn the latent attributes and a decoder network to reconstruct the inputs. However, unlike VAE, LCVA treats the latent attributes as confounders that are assumed to affect both the treatment and the outcome of units. Specifically, given a pair of units u and u, their individual treatment and outcomes, the encoder network of LCVA samples the confounders by conditioning on the observed covariates of u, the treatments of both u and u and the outcome of u. Once inferred, the latent attributes (or confounders) of u captures the spillover effect of u on u. Using a network of users from job training dataset (LaLonde (1986)) and co-purchase dataset from Amazon e-commerce domain, we show that LCVA is significantly more robust than existing methods in capturing spillover effects.
KW - Causal inference
KW - Spillover effect
KW - Variational autoencoder
UR - http://www.scopus.com/inward/record.url?scp=85058040197&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85058040197&partnerID=8YFLogxK
U2 - 10.1145/3269206.3269267
DO - 10.1145/3269206.3269267
M3 - Conference contribution
AN - SCOPUS:85058040197
T3 - International Conference on Information and Knowledge Management, Proceedings
SP - 1679
EP - 1682
BT - CIKM 2018 - Proceedings of the 27th ACM International Conference on Information and Knowledge Management
A2 - Paton, Norman
A2 - Candan, Selcuk
A2 - Wang, Haixun
A2 - Allan, James
A2 - Agrawal, Rakesh
A2 - Labrinidis, Alexandros
A2 - Cuzzocrea, Alfredo
A2 - Zaki, Mohammed
A2 - Srivastava, Divesh
A2 - Broder, Andrei
A2 - Schuster, Assaf
PB - Association for Computing Machinery
T2 - 27th ACM International Conference on Information and Knowledge Management, CIKM 2018
Y2 - 22 October 2018 through 26 October 2018
ER -