1
$\begingroup$

On R's CLVTools package documentation, there's a sentence referring to the pareto/NBD model

I'm working with the BG/NBD model not the pareto/NBD model. I'd like to understand if I can interpret the parameters of this model in a similar way to the documentation on the pareto/NBD model.

From that link above on the Pareto/NBD model:

For the standardPareto/NBD model, we get 4 parameters π‘Ÿ, 𝛼, 𝑠 and 𝛽. where π‘Ÿ, 𝛼 represent the shape and scale parameterof the gamma distribution that determines the purchase rate and 𝑠, 𝛽 of the attrition rate across individualcustomers. π‘Ÿ/𝛼 can be interpreted as the mean purchase and 𝑠/𝛽 as the mean attrition rate

The BG/NBD model I'm working with also returns 4 parameters:

  • r: shape parameter of the Gamma distribution of the purchase process.
  • alpha: scale parameter of the Gamma distribution of the purchase process.
  • a: shape parameter of the Beta distribution of the dropout process.
  • b: shape parameter of the Beta distribution of the dropout process.

The first two of these, π‘Ÿ, 𝛼, are the same as for the Pareto/NBD model. So I'm assuming that I can interpret those two in the same way? "π‘Ÿ/𝛼 can be interpreted as the mean purchase (rate)"?

My question is about the remaining two parameters of the BG/NBD model, a and b. Can I calculate the mean attrition rate in a similar manner by dividing one over the other?

$\endgroup$

1 Answer 1

0
$\begingroup$

In BG/NBD, individual attrition is defined by "p", which follows a beta distribution. The mean of the beta distribution is a/(a+b). Hope this help. BG/NBD was created by Fader, Hardie, and Lee. Please attribute the credit correctly in the future.

$\endgroup$

Your Answer

By clicking β€œPost Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.