Calculate alpha diversity before or after filtering?

Hello,
I have a fundamental conceptual doubt. Once I obtain the metaplan profiles of the samples for further analysis, I filter the species based on an MRA of 0.01 for beta diversity and correlation analysis. However, I am having a hard time deciding when to calculate the number of species observed and Shannon diversity for the samples; should I do it before filtering or after? because I am observing a significant difference before and after filtering.

Thank you.