### Abstract

Four test statistics, namely isotonic test based on arcsin transformation, isotonic test, likelihood ratio test and contrast test, which test the equality of binomial probabilities against ordered alternatives were compared with respect to size and power by extensive simulation. We found that the estimated sizes of the isotonic test based on arcsin transformation and the contrast test overstate the nominal test levels. However, the estimated sizes of the isotonic test and the likelihood ratio test are in close agreement with their corresponding nominal levels. Particularly, when sample sizes are as small as ten, the isotonic test is far better than the likelihood ratio test. For large sample sizes, the former holds little advantage over the latter.

Original language | English |
---|---|

Pages (from-to) | 53-75 |

Number of pages | 23 |

Journal | Journal of Statistical Computation and Simulation |

Volume | 56 |

Issue number | 1 |

Publication status | Published - Dec 1 1996 |

Externally published | Yes |

### Fingerprint

### All Science Journal Classification (ASJC) codes

- Applied Mathematics
- Modelling and Simulation
- Statistics and Probability

### Cite this

*Journal of Statistical Computation and Simulation*,

*56*(1), 53-75.

**A simulation study of several tests of the equality of binomial probabilities against ordered alternatives.** / Kulatunga, D. D Sarath; Asai, Mikiko; Sasabuchi, Shoichi.

Research output: Contribution to journal › Article

*Journal of Statistical Computation and Simulation*, vol. 56, no. 1, pp. 53-75.

}

TY - JOUR

T1 - A simulation study of several tests of the equality of binomial probabilities against ordered alternatives

AU - Kulatunga, D. D Sarath

AU - Asai, Mikiko

AU - Sasabuchi, Shoichi

PY - 1996/12/1

Y1 - 1996/12/1

N2 - Four test statistics, namely isotonic test based on arcsin transformation, isotonic test, likelihood ratio test and contrast test, which test the equality of binomial probabilities against ordered alternatives were compared with respect to size and power by extensive simulation. We found that the estimated sizes of the isotonic test based on arcsin transformation and the contrast test overstate the nominal test levels. However, the estimated sizes of the isotonic test and the likelihood ratio test are in close agreement with their corresponding nominal levels. Particularly, when sample sizes are as small as ten, the isotonic test is far better than the likelihood ratio test. For large sample sizes, the former holds little advantage over the latter.

AB - Four test statistics, namely isotonic test based on arcsin transformation, isotonic test, likelihood ratio test and contrast test, which test the equality of binomial probabilities against ordered alternatives were compared with respect to size and power by extensive simulation. We found that the estimated sizes of the isotonic test based on arcsin transformation and the contrast test overstate the nominal test levels. However, the estimated sizes of the isotonic test and the likelihood ratio test are in close agreement with their corresponding nominal levels. Particularly, when sample sizes are as small as ten, the isotonic test is far better than the likelihood ratio test. For large sample sizes, the former holds little advantage over the latter.

UR - http://www.scopus.com/inward/record.url?scp=0030363427&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0030363427&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:0030363427

VL - 56

SP - 53

EP - 75

JO - Journal of Statistical Computation and Simulation

JF - Journal of Statistical Computation and Simulation

SN - 0094-9655

IS - 1

ER -