IGARCH effect on autoregressive lag length selection and causality tests