PLS Multinomial Logit in Satisfaction Surveys
Satisfaction Surveys collect value judgments of consumers on products or services. The objective of such studies is to prioritize the quality improvements that could be taken ranking them in terms of impact on a Global Satisfaction target variable. The aim of such studies and the data types used-ordinal scales-typically suggest applying Correlations, Multiple Regression or Multinomial Regression to relate global satisfaction appraisal to more specific elements of product quality. However, a common pitfall of such studies is that the various answers of consumers tend to be highly correlated. This leads to well known risks of counter-intuitive results and suppressive effects of actually significant variables. The authors developed a solution based on a PLS Multinomial Regression. The present project is intended to bring to light and measure the benefits of this methodology over the classical approach of Multinomial Regression on a sample often real survey databases selected in various domains. It will also investigate the effectiveness and limits of this methodology in dealing with missing data.
PLS multinomial logistic regression satisfaction
Alain Marechal Paolo Righetti Simon Pioche Hajer Souaifi Riccardo Troso Jean-Pierre Bidon
Interstat Paris GN Research Roma GN Research Paris
国际会议
The 6th International Conference on Partial Least Squares and Related Methods(第六届偏最小二乘及相关方法国际会议)
北京
英文
34-38
2009-09-04（万方平台首次上网日期，不代表论文的发表时间）