I'm running a choice based conjoint ('discrete choice') survey with the primary goal of understanding price elasticity as well as the importance of other attributes. Its my first time working with any sort of conjoint and am receiving mixed advice regarding the best way to analyze the data. 

We are exploring various analysis methods. 

  • A linear model: Utility = intercept + choice bias (includes respondent bias) + M + W + C + T + P + nested effects (e.g., M(P), W(P), etc.) 
  • A logistic model: prob{selection} = (1+exp{-L})-1 , where L=Utility
  • A multinomial logit model: Prob{selection} = exp{-L} / Sum(all possible Exp{L(i) outcomes}, where M, W, etc. are part-worths
  • One advisor prefers the linear model because it controls for bias. Another prefers logit model. The survey software I'm using defaults to logistic regression model. 

    What are the pros/cons of each?

    Similar questions and discussions