Advances and Applications in Statistics
Volume 30, Issue 1, Pages 45 - 75
(September 2012)
|
|
COMPARING CONTINGENCY TABLE TESTS IN LOGISTIC REGRESSION USING BOOTSTRAP SAMPLING
Brian P. Barthel and Mezbahur Rahman
|
Abstract: Different contingency table methods for performing the goodness-of-fit tests have been developed for logistic regression models. Due to complexity in simulating in logistic regression while comparing the performances of these tests, here we consider bootstrap sampling to compare different contingency table methods. We considered five different data sets and all but one showed Pearson’s Chi-square statistic is approximated more closely to the Chi-square distribution compared to others under consideration. Five different test statistics compared are Pearson’s Chi-square test, the Contingency Table Approach, the Hosmer-Lemeshow test and the Rahman and Muraduzzaman [4] test with and without ordering. Although each of the tests seem to be computing the test statistic in the same way, the way in which the data is grouped plays a significant role in the results. |
Keywords and phrases: deviance statistic, maximum likelihood estimate, quantiles. |
|
Number of Downloads: 340 | Number of Views: 800 |
|