If I have several methods to aggregate users ratings towards items (average, median, Bayesian estimate, Dirichlet reputation, .. etc) How can we evaluate the accuracy of items reputations? Now I use the rating prediction method with MAE measure, is there any other method?
Dear Ahmad,
Firstly, there are very little details of what kind of study you are currently pursuing. Hence, my opinion might not reflect the needs of your study. However, you might find the commentary below useful.
There are many different ways to confirm item reputation. This can be done through item analysis that calculates inter-item correlation, mean and standard deviations. You could then use eigenvalues from the factor analysis (if you study requires you to do so) and compare them with the item analysis results. Item difficulty (the sum result of your analysis) will give you the so called item reputation you are looking for.
Rabin
- Badges
- Science topic
- Similar topics
- Comment