Have an idea?

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

What is the effect of a standard error of 0.06 instead of 0.05


when making a CBC design, it is recommended to strive for standard errors for levels of each attribute that are 0.05 or less.

I am wondering now, if my standard errors are 0.06, what is the influence on the outcome of my CBC?

For example: if a product in my simulator has a SoP of 20%.
Does a p-value of 0.05 then mean that I am 95% sure of this SoP of 20%; while a p-value of 0.06 means that I am 94% sure of this SoP?

Many thanks
Kind regards
asked Jun 24 by Tina Van Regenmortel (345 points)

1 Answer

0 votes

The standard errors are not p-values.  It may help to think of them more as margins of error (if you double them).  So a standard error of 0.05 means your utility (not your share) is accurate to +/- 0.10.  Then a standard error of 0.06 means your utility (not your share) is accurate to +/- 0.12.

There's nothing magic about the standard error of 0.05, it's just a guideline about what usually works well in practice.
answered Jun 24 by Keith Chrzan Platinum Sawtooth Software, Inc. (105,750 points)
Hi Keith,
thank you for your answer.
And could you translate this accuracy on the utility to the accuracy on the share?
Is there a +/- % on the share when the accuracy on the utility is +/- 0.12 instead of +/- 0.10?

many thanks
No, there is no simple translation because shares depend on the utilities not only of a given attribute level, but rather on the total utility of an alternative (and the utilities of all the other alternatives).