Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

Prevent overfitting in Latent Class

Hello!

Within my Latent Class I have results for 2-group to 5-group segmentation. In literature, I found that you can make the segment selection based on the Consistent Akaike Info Criterion (CAIC). The lowest CAIC value from all groups should be the segmentation to select. However, it is logical that the more groups you have, the more 'precise' the customer segmentation. How do I prevent overfitting or know that I am overfitting?

For my research, the 5-group identification does not seem to make much sense so therefore I am asking.

Thank you!
asked Jul 30 by vlinderdb (200 points)

1 Answer

0 votes
The goodness of fit statistics (AIC, BIC, CAIC) have built-in penalties for model complexity (in fact, the big difference among these measures is how much they penalize complexity).  They're not like R-squared in a regression analysis that just keeps getting better as you add more parameters.
answered Jul 30 by Keith Chrzan Platinum Sawtooth Software, Inc. (107,050 points)
...