Have an idea?

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

Simulations without a none response

We did a partial profile conjoint for message testing.  Since these are combinations of messages with different themes (attributes), we did NOT have a none option.  It was simply - which combination of messages is most compelling?

We are using the partial vote exp transformation.  I think the answer is simple, but making sure I'm not missing anything.  Without a none option, in our simulator, we would obviously need to simulate more than one message combination/product or else it would always read 100%.  And just because we don't have a none option, the calculations are still the same.  100% gets distributed across whatever products we simulate?  Mathematically, we run the same calculations for all the options we have there - whether there is a none or not?

And the interpretation of the percents are the same - still ratio - a product that gets a 66% is twice as likely to be preferred as the one that is 33%?  And to stat test 2 products, we could do a simple dependent proportions test on these two percents with the sample size?

asked Feb 9, 2016 by stevetlg.com (385 points)

1 Answer

0 votes
Sounds about right.

Please note that a product that gets a share of preference twice as high as another is predicted to be twice as likely to be picked in the context of the CBC questionnaire than the other.  Whether that translates to twice as high to be purchased in the real world is another story that depends on a number of things, including whether the response error to the conjoint questionnaire is similar to the choice error in the real world.
answered Feb 9, 2016 by Bryan Orme Platinum Sawtooth Software, Inc. (144,240 points)