TY - CONF
T1 - Nonparametric analysis of interactions: a review and gap analysis
AU - Conover, William
AU - Salazar-Alvarez, Maria Isabel
AU - Tercero-Gomez, Victor G.
AU - Temblador-Perez, Maria del Carmen
AU - Cordero-Franco, Alvaro E.
PY - 2014
Y1 - 2014
N2 - To improve a system, from a statistical process control approach, tools from the field of design and analysis of experiments might be used. Within this field, experimental and observational studies address the issue of finding causal relationships between potential factors and response variables. To collect data, experimental design techniques are used to obtain information about the different factors of interest. This is followed by an analysis typically performed with ANOVA or regression techniques, where normality of residuals is frequently assumed. Nevertheless, many processes are not prone to having normally distributed errors, outliers might be present, and observations might be heteroscedastic, which diminish the power of detecting effects on the response variable. To address this issue, nonparametric procedures have been developed. However, most techniques are focused on detecting main effects, and only a few of them take into account interactions. This paper discusses the pro
AB - To improve a system, from a statistical process control approach, tools from the field of design and analysis of experiments might be used. Within this field, experimental and observational studies address the issue of finding causal relationships between potential factors and response variables. To collect data, experimental design techniques are used to obtain information about the different factors of interest. This is followed by an analysis typically performed with ANOVA or regression techniques, where normality of residuals is frequently assumed. Nevertheless, many processes are not prone to having normally distributed errors, outliers might be present, and observations might be heteroscedastic, which diminish the power of detecting effects on the response variable. To address this issue, nonparametric procedures have been developed. However, most techniques are focused on detecting main effects, and only a few of them take into account interactions. This paper discusses the pro
M3 - Paper
ER -