An alternative matrix transformation to the F test statistic for clustered data

• Authors: Sukanya Intarapak , Thidaporn Supapakorn

Abstracts

EN

For the regression analysis of clustered data, the error of cluster data violates the independence assumption. Consequently, the test statistic based on the ordinary least square method leads to incorrect inferences. To overcome this issue, the transformation is required to apply to the observations. In this paper we propose an alternative matrix transformation that adjusts the intra-cluster correlation with Householder matrix and apply it to the F test statistic based on generalized least squares procedures for the regression coefficients hypothesis. By Monte Carlo simulations of the balanced and unbalanced data, it is found that the F test statistic based on generalized least squares procedures with Adjusted Householder transformation performs well in terms of the type I error rate and power of the test.

Keywords

EN

adjusted Householder; clustered data; F test statistic; generalized least squares; intra-cluster correlation

• Publisher: Główny Urząd Statystyczny
• Journal: Statistics in Transition new series
• Year: 2019
• Volume: 20
• Issue: 1
• Pages: 153-169

Contributors

author

Sukanya Intarapak

• Department of Mathematics, Faculty of Science, Srinakharinwirot University, Bangkok, Thailand

• https://orcid.org/0000000267709503

author

Thidaporn Supapakorn

• Department of Statistics, Faculty of Science, Kasetsart University, Bangkok, Thailand

• https://orcid.org/0000000300199884

Identifiers

DOI

10.21307/stattrans-2019-009

Biblioteka Nauki

1359200