Desempenho do gráfico de controle CUSUM tabular para o monitoramento da média

• Ano de defesa: 2012
• Autor(a) principal: Lara, Rodrigo Luiz Pereira
• Orientador(a): Ribeiro Junior, José Ivo
• Banca de defesa: Nascimento, Moysés , Faria, Adriana Ferreira de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Estatística Aplicada e Biometria
• Departamento: Estatística Aplicada e Biometria
• País: BR
• Palavras-chave em Português: Qualidade; Poder; Alarme falso
• Palavras-chave em Inglês: Quality; Power; False alarm
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::PROBABILIDADE E ESTATISTICA::ESTATISTICA
• Link de acesso: http://locus.ufv.br/handle/123456789/4048

This study aimed to estimate the probabilities of false and true alarms caused by the Cumulative Sum control chart over the i rational subgroups, according to various combinations of size n of subgroup rational, standardized decision interval h* and the tolerance value k*. To study them, were simulated data from a random variable Y, under normal distribution with mean μ0 = 0 and standard deviation 1 s 0 = for a process under statistical control over 50 rational subgroups, for n between 1-16. To this end, 1000 simulations were performed by rational subgroup. Then four out of control processes have been set for the statistical average at all rational subgroup ( μ1 = μ0 +ds 0 ) in which d refers to the displacement of the average in numbers of standard deviations of the process. The probability of false alarm a decreases with the increase, increase and decrease of k*, h* and i respectively, while the probability of true alarm Pd has direct relation to n and i, and inverse relation to h* for the same pre-defined minimum difference between the means μ0 and μ1 considering the choice of k* as being half of that difference in number of standard deviations. Both probabilities of true and false alarms were obtained by the normal and lognormal 3P probability distributions adjusted to the random variable Z(i) + S (i −1)* H . In order to give a power greater than 0.90 and a equal or less than 0.05 or 0.01 were recommended to different combinations of k*, h*, i and n.