Bias initial values and Use evenly

  • 1
  • Question
  • Updated 9 years ago
  • Answered
What are these two options mean, Bias initial values and Use evenly under create new value expansion?
Photo of Ed

Ed

  • 25 Posts
  • 0 Reply Likes
  • happy

Posted 9 years ago

  • 1
Photo of justin.x.hunter

justin.x.hunter

  • 13 Posts
  • 1 Reply Like
Ed,

Thanks again for your question. Please keep them coming.

In the Value Extension feature, users can expand as many values as they would like into multiple sub-values. If you wanted to get a large cross section of users from multiple U.S. states, for instance, you could group the 50 states into four categories (N, S, E and W) if you wanted. This would likely reduce the number of tests dramatically as compared to listing each of the 50 states as separate values.

Let's say that in the East category, you had NY, NJ, FL, NC, SC, GA, and AL.

What kind of distribution would you like as a test designer? We leave that choice to the user. A test designer who wanted NY to appear a lot more than AL (because , e.g., they have far more NY clients than AL clients) could bias the weighting towards the initial values. That would create a test plan that would consist of a lot more tests with NY in them than with AL in them. If, however, the test designer wanted an approximately equal number of NY mentions as AL mentions, the test designer should use the "Use evenly" option.

Thanks for the question. Please keep them coming.
Photo of Justin Hunter, Hexawise Founder

Justin Hunter, Hexawise Founder, Founder and CEO

  • 246 Posts
  • 15 Reply Likes
Official Response
Ed,

Thanks again for your question. Please keep them coming.

In the Value Extension feature, users can expand as many values as they would like into multiple sub-values. If you wanted to get a large cross section of users from multiple U.S. states, for instance, you could group the 50 states into four categories (N, S, E and W) if you wanted. This would likely reduce the number of tests dramatically as compared to listing each of the 50 states as separate values.

Let's say that in the East category, you had NY, NJ, FL, NC, SC, GA, and AL.

What kind of distribution would you like as a test designer? We leave that choice to the user. A test designer who wanted NY to appear a lot more than AL (because , e.g., they have far more NY clients than AL clients) could bias the weighting towards the initial values. That would create a test plan that would consist of a lot more tests with NY in them than with AL in them. If, however, the test designer wanted an approximately equal number of NY mentions as AL mentions, the test designer should use the "Use evenly" option.

Thanks for the question. Please keep them coming.