Asymptotic power comparison of the chi-square and likelihood ratio tests (Q1084788)

For a multinomial distribution the likelihood ratio test and chi-square tests are used to test simple hypotheses. A Hoeffding-type power comparison is made based on the fact that under an alternative hypothesis the distribution of the test statistic can be approximated by a normal distribution. The theory of large deviations is used to match the significance levels. Three examples are presented to illustrate the theory where in most cases the likelihood ratio test was superior in power.