hi, i'm dealing with a dataset where one of the mo...

# gd-beginnerse

Emile Joubert

06/24/2024, 11:00 AMhi, i'm dealing with a dataset where one of the most important metrics is a binomial proportion with a small sample size (number of successes out of a small number of trials). i'm looking for a metric so that some_metric(p=10, n=10) is closer to 100% than some_metric(p=2, n=2).
is there anything in the MAQL toolbox for a binomial distribution or binomial test? i'm on gooddata cloud.

j

Julius Kos

06/24/2024, 1:27 PMHi Emilie,
Can you please give us more information regarding purpose of using such logic in GoodData context? Could you please give us some “real world” example of using this in our analytics? (input data + desired result)

e

Emile Joubert

06/24/2024, 1:53 PMhi Julius,
the real world use-case is to compare proportions from large and small samples in a way that's sensitive to the difference in certainty. there are rich statistical libraries outside of gooddata, but i don't see how to use those in the context of gooddata after data has been sliced.
to give a concrete example, i want to illustrate a success rate of 2 out of 2 trials alongside a success rate of 98 out of 100 trails, where the number of trials will increase over time. i'm seeing very little support in the MAQL toolbox for a smoothing function that will help me to illustrate this in a way that's sensitive to the difference in confidence about the underlying success rate.

j

Julius Kos

07/02/2024, 12:22 PMHi Emile,
Apologies for my later response. I was double-checking this internally. Currently, MAQL does not include a function for calculating factorials, which are essential for computing binomial coefficients. Factorials tend to produce very large numbers, making them prone to overflow, and not all databases support their direct calculation.
If you wish, please let me know and I will gladly submit a product feedback on your behalf so we can consider how to implement similar functions in our analytics.
In any case, thank you very much for sharing your experience with our product.

5 Views