Inference from plant competition experiments: the effect of spatial covariance