الفهرس 
Power of Goodness of fit Tests for Logistic Regression
ContentsOnly 14 pages are availabe for public view
 |  | from | 119 |  |  |
| | | from | 119 | | |
Abstract
In this study, many GOF tests for logistic regression were discussed and compared to each other from different aspects by a simulation study. The first aspect is the ability of the test to accept the null hypothesis when a true model is fitted, and the second aspect is that the ability of tests to reject the null hypothesis when there was a departure from the true model where the null hypothesis of all tests stated that the model fit. Four different departures were studied: omission linear covariate, quadratic term, interaction term and logarithmic term. The third aspect was to check whether the same tests coded in R under different packages gave the same results or not. Ten different sample sizes were used to show the behavior of each test and the optimum sample size under which a test performed well. An empirical study on breast cancer data were done to enforce results of simulations. Finally, a guide to use GOF tests were proposed not only to those who will use R software and have variety of tests, but also for those who use common tests (Hosmer-Lemeshow test statistics) in another software a guide to these common tests is proposed.