Institutional Repository of Vilnius University
Prekybos platformos pardavėjo metrikų matavimas pirkėjo pusės eksperimentuose /
Abstract
dc:descriptionThis paper explores emerging methodologies for inference in Online Controlled Bipartite Experiments that map interactions between user groups in marketplace settings, allowing us to better quantify the effects of interventions where interference is present. Using real historic experiment data at Vinted, we focus on buyer-seller interactions represented as a bipartite graph, measuring the outcomes as seller metrics. We establish that our in-experiment bipartite graphs can be constructed from various interaction events, and cannot be pre-determined. We evaluate ERL, regression, bootstrapping-based, and covariate-adjusted methods and compare their estimates across two different interaction mappings with two types of metrics, identifying challenges related to their applicability to large-scale data.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Pilkauskaitė, Vaiva,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210643565&prefLang=en_US