Have an idea?

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

Zero Centered Diff


I want to know how to interpret my output. I did a LCA and now, I'd like to investigate my rescaled part-worths for each group. I know that I cannot compare levels from one attribute with levels of another attribute.

But can I e.g. say for group 1:  in attribute colour, the colour red is most preferred (red has highest utility 44.23). Pink (22,20), on the other hand, is less preferred but twice as preferred as Violett (10.71). = Consumers in this group prefer red t-shirts if they're offered.

Plus, may I say for group comparisons:  group 1 and 2 prefer red t-shirts the most. group 3 is more addicted to pink.

In addition, how to interpret the none-option utilities and does negative utility mean 'less preferred or not preferred'.

Thank you!
asked Nov 14, 2017 by Sara

1 Answer

0 votes

Utilities are interval scaled, like Fahrenheit temperature - you can say that 100 degrees is hotter than 50 degrees, but you can't say that it's twice as hot:  the zero isn't a "natural" zero but rather one chosen by convention (e.g. the Celsius scale has its zero at a different place than does Fahrenheit).   So you can talk about more or less utility, but not ratios (like twice as preferred as violet).

For the same reason, you can conclude "more preferred" or "less preferred," but you cannot make an absolute conclusion about "not preferred."  In the example we use in our teaching, if we had an attribute about how much money you'd win in a lottery, the three levels might be $10 million, $20 million and $30 million.  By the standard way of scoring these the $20 million would have a utility in the neighborhood of zero while $30 million would have a large positive utility and $10 million a large negative utility.  This is just an artifact of the coding and of the lack of a natural zero on the utility scale, because I think both of us would be very happy to win $10 million (in fact, if one of us did, then one of us would probably by typing these messages from a warm island somewhere, sipping a cocktail with a little umbrella in it).  

As for "None," its meaning depends on how you worded the none alternative.  If it was "I would wait and buy later," then that's what it means.  If it was "I would go to a different store to shop" or "I would just keep the one I have now," then those are what it would mean.  If you used a more ambiguous "None of these" then your result is correspondingly more ambiguous.  In any case, the none utility becomes a threshold:  if a simulated alternative has higher utility than "None" then it will get more share than None.  If it has a lower utility then it will get less share than None.
answered Nov 15, 2017 by Keith Chrzan Platinum Sawtooth Software, Inc. (103,225 points)