What to do when the P-value of chi-squared is significant but normal inverse Z is not?