On the Reliability of the Classical Chi-Square Test