The efficiency of ordinary least squares in designed experiments subject to spatial or temporal variation