On spurious Granger causality