Banca de DEFESA: FIDEL HENRIQUE FERNANDES

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
DISCENTE : FIDEL HENRIQUE FERNANDES
DATA : 20/02/2018
HORA: 10:00
LOCAL: Auditório do CCET
TÍTULO:

Control Charts for Monitoring of Autoregressive Processes of Integer Values with Inflation or Deflation of Zeros

PALAVRAS-CHAVES:

average run length (ARL); CUSUM chart; deflation of zeros; inflation of zeros; Markov Chain;  statistical process control; Monte Carlo;  Shewhart Chart; ZMGINAR(1) model.

PÁGINAS: 56
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Probabilidade e Estatística
RESUMO:

The time series is a collection of observations sequentially measured over time, being studied with great notoriety in recent years along with statistical process control. The present work aims to study the performance of the CUSUM and Shewhart control charts in the detection of averages of processes with inflation or deflation of zeros through the zero-modified geometric firs-order integer-valued autorregressive process [ZMGINAR(1)]. In this sense, it is also analyzed the sensitivity through simulations, average number of samples exceeding the control limit (ARL), in addition, new estimators were proposed in order to verify, through the Monte Carlo study, the behavior of the estimators through the mean square errors (MSE) and bias in different scenarios, the proposed estimators proved to be more effective. As for the simulation, the different scenarios presented with inflation and deflation of zeros, the CUSUM was more efficient in the scenario with zero deflation and Shewhart with zero inflation in certain cases. In this instance, two applications were considered, one with inflation of zeros and the other with deflation of zeros. Just as in the simulation, CUSUM is best in the zero deflation scenario and the Shewhart with zero inflation. The great differential of this work, is the appearance of deflation of zeros modeled in the control charts, in addition the model to be worked has a known marginal distribution unlike other models, which is an advantage in Implementation and construction of new estimators, in addition to this, it is still considered the parameter estimates by several methods: maximum likelihood, Yule-Walker and probabilit based estimator. 

MEMBROS DA BANCA:
Externo à Instituição - LINDA LEE HO - USP
Interno - 2083593 - LUZ MILENA ZEA FERNANDEZ
Presidente - 1023112 - MARCELO BOURGUIGNON PEREIRA