Statistical evaluation of data from tractor guidance systems

Statistical tools are discussed for the analysis of data collected from tractor guidance systems. The importance of both accuracy and precision is discussed, and statistical tools for analysis are considered which incorporate important features of the data. In particular, accuracy is modelled using a generalized least squares model incorporating autocorrelation, and variances (inverse of precision) using a gamma generalized linear model. The methods are applied to data collected during an experiment conducted with a Trimble receiver used with a Beeline tractor guidance system. Three different scenarios are considered, then compared: a tractor simulating ploughing a field; the tractor pulling a plough with the receivers on the tractor; the tractor pulling a plough with the Trimble receiver on the plough. The change in the precision and accuracy between the scenarios is discussed. Data were recorded over repeated swaths for each scenario. After discussing specific statistical techniques for analysis of this type of data, the collected data are analysed; major conclusions are: The data from the Trimble receiver showed evidence of autocorrelation in the offsets; the plough recorded a variance about three times that recorded by the tractor.