Visual Model Validation via Inline Replication
release_rz4tvv2du5hd3fazii5r2ujije
by
David Gotz, Brandon A. Price, Annie T. Chen
2016
Abstract
Data visualizations typically show retrospective views of an existing dataset
with little or no focus on repeatability. However, consumers of these tools
often use insights gleaned from retrospective visualizations as the basis for
decisions about future events. In this way, visualizations often serve as
visual predictive models despite the fact that they are typically designed to
present historical views of the data. This "visual predictive model" approach,
however, can lead to invalid inferences. In this paper, we describe an approach
to visual model validation called Inline Replication (IR) which, similar to the
cross-validation technique used widely in machine learning, provides a
nonparametric and broadly applicable technique for visual model assessment and
repeatability. This paper describes the overall IR process and outlines how it
can be integrated into both traditional and emerging "big data" visualization
pipelines. Examples are provided showing IR integrated within common
visualization techniques (such as bar charts and linear regression lines) as
well as a more fully-featured visualization system designed for complex
exploratory analysis tasks.
In text/plain
format
Archived Files and Locations
application/pdf 8.4 MB
file_sorgdmpbmvbwjnw4o3xsbsdqm4
|
arxiv.org (repository) web.archive.org (webarchive) |
1605.08749v1
access all versions, variants, and formats of this works (eg, pre-prints)