I have about 10 variables about products. I need to rank order the products by quality. Now, the quality is very subjective thing, so I got all 300 products ranked by domain experts.
Now I have 10 variables, and ranking assigned by an expert. About 1% of products should be excelent, about 10% should be very good, 30% should be good, 30% should be average and the rest is below average.
What sort of modeling technique would you use to make sue of this data and create system which will automatically be able to classify product into those categories and mimic "expert judgement".
Best Answer
The dependent variable is ordinal (excellent > very good > good > average > below average). You could try ordinal logistic regression, with some variable selection method. After choosing the model you could tweak the thresholds to influence proportions falling into various categories.