Notice: the WebSM website has not been updated since the beginning of 2018.

Web Survey Bibliography

Title Evaluation of neural networks to detect suspicious answers in attitude and evaluation research
Year 2002
Access date 24.05.2004
Full text pdf (172k)
Abstract This contribution aims at evaluating neural networks approaches for automatic quality checks, in order to assess their performances under different conditions, and to establish if they are to be preferred to more traditional outliers detection approaches. The attention is specifically directed to attitude and evaluation research, and in particular to categorical ordinal and quantitative discrete measurement scales, which are often used to measure opinions. We carry out a simulation study to compare the proportion of wrong cases introduced in a regular pattern that are detected with alternative approaches. A transform of the received operating characteristic (ROC) curve is used to compare the performances of outliers detection methods based on neural networks with those based on linear and ordinal logistic regression, and to assess their sensitivity to some data characteristics. The neural network approaches are shown to behave not significantly better than simpler diagnostic methods in a range of common situations.
Access/Direct link Homepage - conference (full text)
Year of publication2002
Bibliographic typeConferences, workshops, tutorials, presentations
Print

Web survey bibliography - Italy (57)

Page:
  • 1
  • 2
Page:
  • 1
  • 2