Analysis of covariance for the removal of temporal effects in experiments with serial treatments