Framework of decision in data modeling for quality improvement

Purpose – The purpose of this paper is to propose a framework of decision making to aid practitioners in modeling and optimization experimental data for improvement quality of industrial processes, reinforcing idea that planning and conducting data modeling are as important as formal analysis. Design/methodology/approach – The paper presents an application was carried out about the modeling of experimental data at mining company, with support at Catholic University from partnership projects. The literature seems to be more focussed on the data analysis than on providing a sequence of operational steps or decision support which would lead to the best regression model given for the problem that researcher is confronted with. The authors use the concept of statistical regression technique called generalized linear models. Findings – The authors analyze the relevant case study in mining company, based on best statistical regression models. Starting from this analysis, the results of the industrial case study illustrates the strong relationship of the improvement process with the presented framework approach into practice. Moreover, the case study consolidating a fundamental advantage of regression models: modeling guided provides more knowledge about products, processes and technologies, even in unsuccessful case studies. Research limitations/implications – The study advances in regression model for data modeling are applicable in several types of industrial processes and phenomena random. It is possible to find unsuccessful data modeling due to lack of knowledge of statistical technique. Originality/value – An essential point is that the study is based on the feedback from practitioners and industrial managers, which makes the analyses and conclusions from practical points of view, without relevant theoretical knowledge of relationship among the process variables. Regression model has its own characteristics related to response variable and factors, and misspecification of the regression model or their components can yield inappropriate inferences and erroneous experimental results.