### Abstract

This manuscript overviews exact testing of goodness of fit for log-linear models using the R package exactLoglinTest. This package evaluates model fit for Poisson log-linear models by conditioning on minimal sufficient statistics to remove nuisance parameters. A Monte Carlo algorithm is proposed to estimate P values from the resulting conditional distribution. In particular, this package implements a sequentially rounded normal approximation and importance sampling to approximate probabilities from the conditional distribution. Usually, this results in a high percentage of valid samples. However, in instances where this is not the case, a Metropolis Hastings algorithm can be implemented that makes more localized jumps within the reference set. The manuscript details how some conditional tests for binomial logit models can also be viewed as conditional Poisson log-linear models and hence can be performed via exactLoglinTest. A diverse battery of examples is considered to highlight use, features and extensions of the software. Notably, potential extensions to evaluating disclosure risk are also considered.

Original language | English (US) |
---|---|

Pages (from-to) | 1-17 |

Number of pages | 17 |

Journal | Journal of Statistical Software |

Volume | 17 |

Issue number | 7 |

State | Published - 2006 |

### Fingerprint

### Keywords

- Exact testing
- Fisher's exact test
- Hypothesis testing
- Poisson models

### ASJC Scopus subject areas

- Software
- Statistics and Probability

### Cite this

**Exact hypothesis tests for log-linear models with exactLoglinTest.** / Caffo, Brian S.

Research output: Contribution to journal › Article

*Journal of Statistical Software*, vol. 17, no. 7, pp. 1-17.

}

TY - JOUR

T1 - Exact hypothesis tests for log-linear models with exactLoglinTest

AU - Caffo, Brian S

PY - 2006

Y1 - 2006

N2 - This manuscript overviews exact testing of goodness of fit for log-linear models using the R package exactLoglinTest. This package evaluates model fit for Poisson log-linear models by conditioning on minimal sufficient statistics to remove nuisance parameters. A Monte Carlo algorithm is proposed to estimate P values from the resulting conditional distribution. In particular, this package implements a sequentially rounded normal approximation and importance sampling to approximate probabilities from the conditional distribution. Usually, this results in a high percentage of valid samples. However, in instances where this is not the case, a Metropolis Hastings algorithm can be implemented that makes more localized jumps within the reference set. The manuscript details how some conditional tests for binomial logit models can also be viewed as conditional Poisson log-linear models and hence can be performed via exactLoglinTest. A diverse battery of examples is considered to highlight use, features and extensions of the software. Notably, potential extensions to evaluating disclosure risk are also considered.

AB - This manuscript overviews exact testing of goodness of fit for log-linear models using the R package exactLoglinTest. This package evaluates model fit for Poisson log-linear models by conditioning on minimal sufficient statistics to remove nuisance parameters. A Monte Carlo algorithm is proposed to estimate P values from the resulting conditional distribution. In particular, this package implements a sequentially rounded normal approximation and importance sampling to approximate probabilities from the conditional distribution. Usually, this results in a high percentage of valid samples. However, in instances where this is not the case, a Metropolis Hastings algorithm can be implemented that makes more localized jumps within the reference set. The manuscript details how some conditional tests for binomial logit models can also be viewed as conditional Poisson log-linear models and hence can be performed via exactLoglinTest. A diverse battery of examples is considered to highlight use, features and extensions of the software. Notably, potential extensions to evaluating disclosure risk are also considered.

KW - Exact testing

KW - Fisher's exact test

KW - Hypothesis testing

KW - Poisson models

UR - http://www.scopus.com/inward/record.url?scp=33845355651&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33845355651&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:33845355651

VL - 17

SP - 1

EP - 17

JO - Journal of Statistical Software

JF - Journal of Statistical Software

SN - 1548-7660

IS - 7

ER -