Have an idea?

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

Interpretation of the Effects of a Logit-Output

Dear Sawtooth Community,

I have a question about the "effects" in the output of a CBCA logit analysis.

After reading and summarizing the literature on Logit, I understand these in general. However, I wonder how I interpret the effects. According to https://www.sawtoothsoftware.com/download/techpap/cbctech.pdf you have to exponentiate the effects for a comparison. Can I therefore assume that these are the logit coefficients that have already been multiplied by the independent variable?
If this is not so, I would be glad if someone could explain to me what the stated effects says.

Best regards,

asked Jun 30, 2018 by anonymous

1 Answer

0 votes
The effects are the CBC utilities.  They are NOT the raw logit coefficients - for a 4-level attribute there would be 3 coefficients in the model (because of the effects coding of the categorical attributes).  In showing 4 effects for a 4-level attribute we've taken the 3 logit coefficients and multiplied them by the attribute's coding scheme (so that you have one effect per attribute level, for categorical attributes).
answered Jul 1, 2018 by Keith Chrzan Platinum Sawtooth Software, Inc. (105,750 points)