Browse
Recent Submissions
Item Improvements In Statistical Monitoring Methods Using Modified Successive Sampling(UMT, Lahore, 2024) Mehvish HyderIn practice, the Statistical Process Control (SPC) toolkit is used to detect the shift in process's location and dispersion parameters. In this toolkit, the control charts are the most frequently used and effective tool for the real-time surveillance of a process. A control chart indicates the process's situation, whether in-control or out-of-control, due to special cause variation in the process. In addition, for the quality assessment, samples are generated through simple random sampling; however, to reduce the sampling time and cost, samples are preferably generated through the modified successive sampling (MSS) mechanism. The existing memoryless (Shewhart) control charts based on MSS scheme for location and dispersion parameters monitoring have very poor performance, and not a single study has made on memory-type control charts based on MSS technique. There was a big research gap presents in memory-type control charts under MSS scheme for the monitoring of both location as well as dispersion parameters and this was an open problem which needed to be fill. Thus, by overcoming this problem, we propose some new efficient and cost-effective memory-type control charts based on MSS technique to detect the shift in location and scale parameters in this thesis.Item DEVELOPMENT OF HYBRID CLASSIFIERS FOR CLASSIFICATION OF HIGH DIMENSIONAL DATA WITH LIMITED SAMPLE SIZE(UMT, Lahore, 2024) Arzoo KanwalClassification is one of the potential and widely applied domain of statistical modeling. Classifiers have a wide variety of applications in high dimensional data sets which include areas of data mining, natural language processing, finance, voice recognition, signal decoding, medicine, chemometrics, etc. Multicollinearity, Heterogeneity and Non-normality have an impact on how well classifiers work in data sets with many dimensions and tiny sample numbers. The existing hybrid high dimensional classifiers do not work optimally under these conditions. In this work, new Hybrid Classifiers are developed which are robust under the aforementioned assumptions. The consistency and classification accuracy of the estimations are unclear when dealing with such high-dimensional data. Assuming that the multivariate regular distribution of the PLS scores, linear discriminant analysis is typically used in conjunction with PLS scores.Item NEW ESTIMATION METHODS TO HANDLE ULTICOLLINEARITY AND DISPERSION ISSUES IN CONWAY MAXWELL’S POISSON REGRESSIONMODELLING(UMT, Lahore, 2024) Faiza SamiThe count regression model is widely used in real life. The model is less precise in the presence of multicollinearity and dispersion. The flexible model with biased estimation method is an appropriate. The Conway-Maxwell Poisson Regression model (COMPRM) is used to handle dispersion issues and biased estimation methods resolve the problem of multicollinearity. COMPRM can resolve the problem of over, under and equi-dispersion cases by using additional dispersion parameter. As, COMPRM uses mean and dispersion parameter in modelling by using two GLM. In this study, we use mean model for estimation with all three cases of dispersion. Beside this, biased estimation methods provide an estimator that give efficient performance as compared to traditional method of estimation. To accomplish the objectives of the study we conduct simulation study under various conditions and two real life applications. We proposed a new generalized ridge estimator, generalized liu estimator, generalized modified one parameter liu estimator and generalized almost unbiased ridge estimator for COMPRM.