Graduate students Payton Miloser and Rebekah Scott placed first and second in the Graduate Student Oral Competition at the 2025 Conference on Applied Statistics in Agriculture and Natural Resources (CASANR), held May 15 at the University of Florida.
First Place: Payton Miloser
Measuring Spatial Consistency in Simulated Crop Yield with Tensor-Product Splines and Interclass Correlation
Authors: Payton Miloser¹, Philip Dixon²
Abstract: Crop yield is spatially consistent when the high-yield areas of a field (or low-yield areas) occur in the same places from year to year. Using precision ag data from previous years to design a management plan for the current year assumes there is spatial consistency, but there are few statistical methods to assess this consistency, especially for data from more than two years. Generating simulated field data from a Gaussian process, we show the use of generalized additive models with tensor-product splines on a range of consistent to inconsistent spatial trends across multiple years of data. This avoids pairwise comparisons, expanding on current methodology used for precision ag data. Measuring this consistency is as important as identifying it, thus using interclass correlation coefficients we have the ability to detect small amounts of inconsistencies across the field and locate those for improvements in farming effectiveness.
Second Place: Rebekah Scott
Using Segmented Regression to Estimate Decay of Curve in Seismic Data
Abstract: In a recent statistical consultation, we assisted a geology graduate student in interpreting simulated seismic data. The problem was to estimate the decay of the curve, the relationship between the log frequency and amplitude after an abrupt change called corner frequency, in seismic waves at nine observations points along a simulated slip fault. We did this using segmented regression. Some of the issues we had to deal with were deciding between a separate or combined analysis for the observation points, determining the source of larger variability in the central observation point, and helping the client interpret the coefficients of this non-linear model. Estimating the decay of the curve in the simulated data allowed our client to validate the simulation and compare the results to current seismic theory.
This presentation was part of collaborative consulting work with Elizabeth Sunday in the Department of Earth, Atmosphere, and Climate.
The Department of Statistics is proud of Payton and Rebekah’s outstanding work.