Dear Dr. Orme,
I deeply appreciate your response.
The number of CVA questions included in the study was only 1.5 times more than the parameters that needs to be estimated, so I do not have enough degrees of freedom. Still, I did get rid of those individuals with low average r square (0.09) and reran the data and the average r square value increased by 0.06. Would 0.19 to 0.25 be much different?
I've read your first book and I could not find the exact source to site what would be the good threshold of the average or square value to cite in my paper.
Would the answers of yours and Dr. Keith that posted under the sawtooth forum be usuable?
Would there be a reference to reflect the model fit (r squared)? MAE (4.XX%)? Hit rate (more than 60-75%)?
Is MAE in conjoint equivalent to MAPE?
I removed all the constraints, as it makes more sense, and the outcome was a slightly better result in terms of the average r square, from 0.25 to 0.30, but with about the same MAE and even worse hit rate.
My field of study is social science.
When running typical regression, R squared value of .3, representing the variability explained by the variables, would be considered as small effects.
However, if I'm understanding this correct, the R squared value for conjoint model which represents the correlation between the estimated and the actual preferences. The value should be at least .6 or more and if not, it suggests my model fit is quite bad. Correct?
Thank you so much for your feedback and insights.