Have an idea?

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

Calculate Zero-centered utilties for all iterations in alpha file

Hello everyone,

I have seen the instruction for transforming raw data into ZCD, but I get stuck at the second Step:

"2. Then, for each attribute compute the difference between best and worst utilities.  Sum those across attributes."

What exactly do you mean with sum those across attributes? I have 34 variables, price(linear), color (4 levels), Label (4 levels), pricexlabel (4Level), pricexcolor (4Level), colorxlabel (16Label)+None option. I have now 7 values of (MAX-MIN) in every row/iteration (the ones for price and the None option are Zero). Do I sum them up horizontally AND vertically to receive one sum?

"3.  Take 100 x #attributes and divide it by the sum achieved in step 2.  This is a single multiplier that you use in step 4."

So basically The sum/(34*100) =X

"4.  Multiply all utilities from step 1 by the multiplier.  Now, the average difference between best and worst utilities per attribute is 100 utility points."

X* each an every utility I find in the alpha file? Do I do it differently with linear attributes?
I assumed I could go the other way round (since I have the average zcd for the utility report) but they don't all have the same multiplyer.

I hope you understand where I am struggeling and hopfeully you could help me.

Best regards
asked Jun 13, 2017 by JanaJardir2 (235 points)
retagged Jun 13, 2017 by Walter Williams

1 Answer

0 votes
If you are doing something by hand, I would first start with a "plain" model, i.e. nothing fancy.  Everything set to part-worth and no interaction effects.  That should help make sure you are doing things correctly and you should be able to match what Lighthouse Studio creates for the rescaled utilities.

Rescaling is done per respondent, so you would calculate the min and max per attribute for each respondent to find the range of each attribute for each respondent.  Interaction effects are not included in the range, since they are modifying parameters that are only in use for combinations of levels across attribute, and importance scores are calculate per attribute.  For linear attributes, you typically multiply them by the high and low values so you can get a range.  The none option is not included because it does not have a range, but is a fixed value.

You could rescale the alphas file if you want, but that is the upper level of the model.  Rescaling is typically done on an individual level to make the utilities of respondent 1 more comparable to the utilities of respondent 2, etc.
answered Jun 13, 2017 by Brian McEwan Gold Sawtooth Software, Inc. (37,710 points)