On-farm experimentation (OFE) has allowed farmers to improve crop and land management and increase their productivity over the years. One of the most prevalent OFE designs has been the randomized complete block design (RCBD) with full field length strips as individual plots, replicated a minimum of three-time. This design, commonly referred to as on-farm strip trials pose challenges for both farmers and scientists, due, among others, to intense planning requirements and limited statistical power. Single-strip evaluation trials are easier to implement for farmers, allowing more farmers to participate in on-farm research. Lack of replication of strips with a single-strip design has limited use of results but with growing availably of yield monitor systems that collect yield data every second that a harvester is in the field, re-evaluation of statistical approached to evaluating such designs is needed. The objective of this research, therefore, was to build a statistically appropriate framework for analyzing single-strip trials for improved OFE for both scientists and farmers. Analyzing a single-strip treatment trial poses challenges due to spatial and temporal yield variability that could potentially interact with the treatment alternative being analyzed. To control for temporal variability, multiple years of geo-referenced yield data need to be analyzed. The first chapter of the thesis explores the three most commonly-used spatial estimation methods, namely nearest neighbor (NN), inverse distance weighting (IDW), and kriging, for converting point data gathered with yield monitors to regular, grid-based, raster maps, which are necessary for temporal analysis of yield. Seven spatial estimation methods (NN, IDW using 10, 20, 30 and all data points and kriging with exponential and Matérn covariance functions) were evaluated to determine the method that most accurately captures intra-field spatial variability of corn silage and corn grain yield in New York. Normalized root means squared error (NRMSE) was used to evaluate the accuracy of the spatial estimation methods. Kriging with the Matérn covariance function resulted in the most accurate corn silage and grain yield raster maps at both the farm and field levels. There were statistically significant differences in NRMSE between kriging with the Matérn isotropic covariance function and all other models for both corn silage and grain, regardless of field size, year, timing of data collection, or farm that supplied the data. While it is known that multiple years of yield maps are needed to delineate farm-specific management zones, the effect of number of years of data on the management zone delineation has not been studied. The second chapter explores this topic. Multiple years of corn silage and grain yield were analyzed to calculate average temporal yield and yield variability, as impacted by the number of years of yield data included for the computation. Results showed that data should first be examined for the presence of a yield trend when estimating average yield. Including only the most recent four to five years of data reduced the risk of misrepresenting the expected average yield when there was a yield trend. Years that had extreme weather could greatly impact average yield measurements for farms with fewer years of data. The temporal standard deviation in yield was most consistent when all years of data were included. We concluded that farms interested in developing yield stability management zones should use at least four to five years of yield data and continue to add new years of data for improved delineation of zones over time. The third chapter explores statistical frameworks for estimating the effect of the treatment from a single-strip OFE, using georeferenced yield monitor data and a historic yield record of the farm. This study analyzed data from single strip-treatment trials on six site-years in 2018 and 2019 for two farms located in Central New York. We examined two approaches, namely Least Squares and Generalized Least Squares with spatial covariance, for estimating the effect of the treatment, and two approaches, with the estimation assuming independence and spatial covariance, for estimating the standard errors. Results suggested Least Squares approach should be used for treatment effect estimation, while spatial covariance should be assumed when estimating the standard errors for single strip treatment trials with high resolution spatial yield data, historic yield monitor data, and yield stability-based management zones. This single-strip spatial evaluation approach allows for more field trial data to be added over time for a better understanding of drivers for outcomes such as yield and the need for site-specific management in unstable yielding zones.