![]() | Only 14 pages are availabe for public view |
Abstract This study presents a new proposed method used to the residual control charts (RCCs) on some regression models, such as Poisson, negative binomial, logistic, and gamma regression, after resolving multicollinearity problems by utilizing the ridge technique. ARL is selected as the measurement for the evaluation of the RCCs. Simulated data and application data using real data on water quality. We conclude from the results that the best performances of charts are k_1,〖k_2,k〗_3,〖and k〗_4 for real data, but in simulation studies they are k_3,k_4,and k_5 for two types. So the best performance effect of the residual control chart is k_5 for two types of residuals used in negative binomial and Poisson. The results show us that the best performance of the residual-based Shewhart control chart for two types of residuals in a simulation study is k_1,〖 k〗_2,k_3,k_5, and k_4 but the best performance of the real study is k_1,k_2, k_3, and k_4. So the best effect of the performance of the residual-based Shewhart control chart in our two cases (simulation and real study) are k_3 and k_4. We also note that the residual values are close to each other in the two types after being processed by the ridge method, and in logistic regression, the best ridge estimators are k_1,〖k_2 ,k〗_3,k_4 and they are all from the control chart. There is no change (e.g., in-control) and the control chart for ARL is k_3 in ordinary raw, but in Pearson the ARL is k_4 and〖 k〗_3. Furthermore,〖 k〗_2,k_3,k_5 are the best k-Ridge parameters. But in gamma, the best performance of k’s of Average Run Length (〖ARL〗_0) for ordinary raw residual is k_1=1.834 and the value of the best performance of 〖ARL〗_1 is k_3 and k_(5 )is equal to 0.8151. However, in the Pearson Residual, the best performance of 〖ARL〗_0 is equal tok_4 . The best performance of 〖ARL〗_1is k_5 and it is equal to 0.815. And we also found that the results corresponding to two types of residual values are consistent after being processed by the ridge method and the joint-charts show the samples to be at out-of-control limits for count data. |