|
Activity Number:
|
52
- Statistical Process Control
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Quality and Productivity Section
|
|
Abstract #309812
|
|
|
Title:
|
Multivariate Semiparametric Control Charts for Mixed-Type Data
|
|
Author(s):
|
Elisavet Sofikitou* and Marianthi Markatou and Markos Koutras
|
|
Companies:
|
University at Buffalo, Department of Biostatistics, USA and University at Buffalo and University of Piraeus, Department of Statistics and Insurance Science, Greece
|
|
Keywords:
|
Artificial Intelligence;
Average Run Length;
Clustering;
False Alarm Rate;
KAMILA Algorithm;
Kernel Density Estimation
|
|
Abstract:
|
We propose a class of multivariate control charts for mixed-type data, based on modified versions of the KAMILA clustering method. Using the classical non-parametric set up, we assume that there are two clusters: one representing the reference sample and the other the test sample. For the continuous variables, the distances to the nearest centroids are exploited in order to construct an appropriate univariate kernel density estimator, while the categorical characteristics are considered to be multinomial random variables. We propose several algorithms for establishing control charts for mixed-type data, study the characteristics of them as well as other theoretical aspects related to the algorithmic procedures employed. We provide tables with some typical in-control values of the False Alarm Rate and the Average Run Length for the implementation of our charts. The performance of the new charts is compared to that of other techniques proposed recently for mixed-type data. The simulation study (carried out under different scenarios) reveals that the suggested method outperforms the existing ones in various cases. The applicability of the new schemes is demonstrated using real-data.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2020 program
|
