Yazar "Derici, Didem" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • [ X ]
    Öğe
    Investigation of Factors Affecting the Performance of Poisson Regression Model: A Simulation Study
    (2021) Derici, Didem; Temel, Gülhan Orekici; Kaya, Irem Ersöz
    Objective: The aim of this study is to compare the performance of the regression correlation coefficient (RCC) and its estimators, which is the measure of the power of the regression model, in terms of bias and root mean square error (RMSE), considering the multicollinearity for different sample sizes, regression intercept (α) parameters and missing observation rates. Material and Methods: Data were produced with MATLAB for the combination of different sample sizes (n: 50, 100, 200, 500), correlations between independent variables (r: 0.0, 0.70), α parameters (α: 0.0, 0.80) and missing observation ratios (m: 0%, 5%, 10%, 20%) for estimators of RCC. Then, randomly subtracted data were analyzed with Poisson regression model and this process was repeated 1,000 times. Results: Simulation results showed that leave-one-out-cross validation estimator ( crs) had the smallest bias and RMSE compared with the other estimators for n=50. All estimators provided similar bias and RMSE values depending on the increase in sample size. It was observed thatwas more robust than the others against to missing ratio for small sample size. In contrary to missing ratio, the most adversely affected estimator by multicollinearity was. Conclusion: The usage of Poisson regression model in clinical trials is wide spreading. Therefore, it is important to evaluate the success of the model in the best way. The study was the first study which investigated the effect of missing ratio on the estimators of RCC for Poisson regression model. It was detected thatwas the most successful estimator in the case of missing observations.