DIY MaxDiff

From Q
Jump to: navigation, search

The files below are for a MaxDiff webinar run on the 31st of May 2017.

Video

 


NOTE: The function used in this video has changed slightly since the recording. The outputs at 24:12 will now look different when taking the same steps as in the video:

  1. The current output shows coefficients (aka utilities). Previously we showed shares. If you look at the ordering of the results they are the same. Google is first, etc.
  2. We changed to showing utilities as sometimes when you show shares for the less liked alternatives they all show at 0, which is not so useful. Also, when comparing multiple classes (segments) showing coefficients is much more useful.
  3. To get the shares, select the output and then click Create > Marketing > MaxDiff > Diagnostic > Class Preference Shares.
  4. Jargon. "Latent Class Analysis" in this case is shorthand for "Latent Class Multinomial Logit". A Latent Class Multinomial Logit Model with only 1 class is most commonly known as just Multinomial Logit. The two headings, while they look different, are really saying the same thing.

Slides

The slides are on SlideShare

Data

Designs

Further reading: MaxDiff software