lgingerich commented on issue #1545: URL: https://github.com/apache/datafusion/issues/1545#issuecomment-4651036737
I was taking a look into this and spent a bit of time on the algorithmic side of it to compare numerical accuracy of Welford+float, Welford+decimal, and sum-of-squares + decimal. TLDR is that it seems that Welford+float is always better than attempting to calculate the variance (and therefore std dev) natively for decimals due to how truncation/rounding error accumulates. Reference: https://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Welford's_online_algorithm Here's a simple example of variance error accumulation for a set of random inputs: <img width="2401" height="1177" alt="Image" src="https://github.com/user-attachments/assets/41f67276-5e6d-498e-9e26-89dad5d36249"; /> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
