I am interested in utilizing Maximum Difference Scaling (Max Diff) for a survey, but apart from the Sawtooth software, I was not able to find a good answer on how to best analyze Max Diff. What is the best way to represent the result in my dataset, and which steps are used to analyze the data? I would prefer to use R or SPSS for analysis.
- Anybody can ask a question
- Anybody can answer
- The best answers are voted up and rise to the top
Estimation of 'counts analysis' of Max-Diff data in both R and SPSS is straightforward (after recoding it is just computed as an average). This page explains the how to do such analyses.
More complicated analyses proceed by either using a 'tricked' logit model (e.g., Sawtooth Software), or, use the rank-ordered logit model (see Allison, P. D. and N. A. Christakis (1994). "Logit Models for Sets of Ranked Items." Sociological Methodology 24: 199-228), which is available in SAS (using PHREG with the EXACT option for ties) or in Q using a 'Question Type' of Ranking.
This page shows a case study which shows how to compute the counts, rank-ordered logit with ties and tricked logit models in R.