Analysis of a Series of Long-Term Grassland Experiments with Autocorrelated Errors

Publisher Website
Google Scholar
Abstract
The residuals in perennial crop experiments are autocorrelated. The autocorrelation normally decreases as the time interval increases. This error structure can sometimes be adequately represented by an autoregressive moving average model, ARMA (p, q). The problem of identifying and estimating the ARMA (p, q) error models for experimental error differs in some aspects from the common time series applications, and an error component independent of time differences, plot error, may also be present. The errors in a group of experiments were identified as 2nd order autoregressive, AR(2). A transformation is given for transforming data with AR(2) residuals prior to the regression analysis of the vectors of annual treatment responses.

This publication has 0 references indexed in Scilit: