|
Activity Number:
|
243
- Statistics in Sports and Beyond
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistics in Sports
|
|
Abstract #318120
|
|
|
Title:
|
Comprehensive Geostatistical Techniques Illustrated by 'Spatial Boxplots' of Baseball Heat Maps
|
|
Author(s):
|
Dana Sylvan* and Jared Cross
|
|
Companies:
|
CUNY Hunter College and St Ann's School
|
|
Keywords:
|
heat maps;
spatial prediction;
visualization
|
|
Abstract:
|
Spatial processes have been increasingly analyzed in recent years due in part to the explosive growth and affordability of computing capabilities. Many fields of applied research employ geostatistical methods to analyze spatial patterns in data with complex structures. Inspired by the pioneering work of John Tukey, we adapt the five-number summary concept to create a visual tool that can be construed as a spatial analog of the boxplot and illustrate it on baseball data. We use the freely available Statcast Trackman data which provides continuous location coordinates for individual baseball pitches using Doppler radar. In this study we focus on Major League Baseball pitches from 2006 to 2018. The stochastic process underlying batting ability is assumed to be a spatial Gaussian field with isotropic covariance that is estimated from the aforementioned data. We then use Kriging Residuals, the winning algorithm among the several procedures introduced in Cross and Sylvan (2015) to obtain best estimates of heat maps of batting ability for individual players. We also assess uncertainty in these estimates by using simulations and resampling and show “spatial boxplots” for visualization.
|
Authors who are presenting talks have a * after their name.
