A note on recognizing autocorrelation and using autoregression

The presence of autocorrelation in the analysis of a variable sampled sequentially at regular time intervals appears to be unknown to many agricultural meteorologists despite abundant documentation found in the traditional meteorological and statistical literature. It follows that the statistical consequences as well as methodological alternatives are also unknown. Through an example using paired radiometer observations, this note discusses recognition of autocorrelation as well as the importance of testing ordinary least squares regression parameters in the presence of autocorrelated residuals. An autoregression example is presented as one alternative way to analyze the given dataset.