TY - JOUR
T1 - Modified Tukey's control chart
AU - Tercero-Gomez, Victor
AU - Ramirez-Galindo, Jose
AU - Cordero-Franco, Alavarado
AU - Smith, Milton
AU - Beruvides, Mario
PY - 2012
Y1 - 2012
N2 - Phase I of control analysis requires large amount of data to fit a distribution and estimate the corresponding parameters of the process under study. However, when only individual observations are available, and no a priori knowledge exists, the presence of outliers can bias the analysis. A relatively recent and successful approach to address this situation is Tukey's Control Chart (TCC), a charting method that applies the Box Plot technique to estimate the control limits. This procedure has proven to be effective for symmetric distributions. However, when skewness is present the average run length performance diminishes significantly. This article proposes a modified version of TCC to consider skewness with minimum assumptions on the underlying distribution of observations. Using theoretical results and Monte Carlo simulation, the modified TCC is tested over several distributions proving a better representation of skewed populations, even in cases when only a limited number of observations are available.
AB - Phase I of control analysis requires large amount of data to fit a distribution and estimate the corresponding parameters of the process under study. However, when only individual observations are available, and no a priori knowledge exists, the presence of outliers can bias the analysis. A relatively recent and successful approach to address this situation is Tukey's Control Chart (TCC), a charting method that applies the Box Plot technique to estimate the control limits. This procedure has proven to be effective for symmetric distributions. However, when skewness is present the average run length performance diminishes significantly. This article proposes a modified version of TCC to consider skewness with minimum assumptions on the underlying distribution of observations. Using theoretical results and Monte Carlo simulation, the modified TCC is tested over several distributions proving a better representation of skewed populations, even in cases when only a limited number of observations are available.
KW - Average run length
KW - Box plot
KW - Skewness
KW - Tukey's Control Chart
UR - http://www.scopus.com/inward/record.url?scp=84862860621&partnerID=8YFLogxK
U2 - 10.1080/03610918.2011.606951
DO - 10.1080/03610918.2011.606951
M3 - Article
AN - SCOPUS:84862860621
SN - 0361-0918
VL - 41
SP - 1566
EP - 1579
JO - Communications in Statistics: Simulation and Computation
JF - Communications in Statistics: Simulation and Computation
IS - 9
ER -