Have an idea?

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

Affects of Covariates and cryptic error

Thank Bryan

I checked the paper and I am sorry if I got it wrong but I believe I have a rather different situation. in my ACBC study without BYO, the covariates that are included consists of likert questions with 3 or 4 items. So, i am checking how a person being individualist or collectivist affects her choices on attributes by asking 4 questions. I want to know how a person being more individual make differences on the choice compared to the collectivist one.

I computed a new variable on SPSS by having means of the 3 items to see if a person is individualist or not and I wanted to add it to my ACBC data with a new column and then test it, as I am using SSI web analysis module. Adding column was not possible actually as I have an error that says ''Object reference not set to an instance of an object" which is also another question I wanted to ask.

Therefore, I wanted to know if I am following the right approach to see the affect of my covariates on attributes or this approach is more related to traditional CBC and segmentation analysis. Also, I would be glad if you could help me with the solution to the "Object reference not set to an instance of an object" error.

Thank you !
related to an answer for: Affects of covariates on attributes (ACBC)
asked Dec 10, 2014 by Mir (400 points)
retagged Dec 10, 2014 by Walter Williams

1 Answer

0 votes
I think perhaps I was misunderstanding.  I thought you had applied your Likert scale variable questions as covariates within the HB estimation routine, meaning that the alpha file would contain successive estimates across iterations of the effect of the covariates (as regression weights) upon the part-worth utilities.

But, maybe you did not include these Likert questions during your HB estimation, but only need to use them as predictors of part-worth utilities estimated separately using plain vanilla HB (without covariates)?

There are many ways to do that, but I would think you'd use the zero-centered diffs (normalized) utilities that our HB software outputs.  You'd read those along with your Likert scale questions into a software package like SPSS, R, SAS, etc.  Then, you'd do statistical tests on whether the Likert scale questions were significant predictors of respondents' preferences.   One simple test flows from regression, for example.
answered Dec 10, 2014 by Bryan Orme Platinum Sawtooth Software, Inc. (131,890 points)
ACBC covariate estimation - alpha file