Have an idea?

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

claculation of degrees of freedom

Hi,

how can I calculate the degrees of freedom in order to check if Chi Square for latent class (CBC) is sig.

In my study I included 6 attributes with 3 level each (in total 18 levels).

Thanks and best,
Judith
asked Jun 11, 2015 by J_henr03 (190 points)
Judith,

To extend this to the latent class case you would also have to account for the number of segments, multiplying by the number of segments minus one.  So with 6 attributes 3 levels each and a none with 4 segments you'd have (18-6+1)*3 = 39 parameters.

1 Answer

0 votes
The number of parameters estimated is as follows for a conjoint study:

If you are using the default (part-worth utility estimation),

Total#_Levels-Total#Attributes

And, if you have a None choice, add 1 to that.

For example, if you have 6 attributes each with 3 levels plus a None, the total number of parameters estimated is:

18-6+1=13
answered Jun 11, 2015 by Bryan Orme Platinum Sawtooth Software, Inc. (140,065 points)
But, if you add "interaction terms" to the model beyond the main effects, those add (j-1)(k-1) parameters to the model, where j is the number of levels in the first attribute of the interaction effect and k is the number of levels in the second attribute.
I did not include a none choice option. Does that mean my dgrees of freedom are 18-6=13?
(6 attributes with 3 levels each, total 18 levels)
...