dist_calc_seq binary jaccard calculation

[scilexenko]

Dear David,

I'm diving into the beta-volatility calculations, and would like to clarify some things. I used dist_calc_seq to calculate the difference in Jaccard dissimilarity between two time points. After some deduction and your function analysis I came up with next formulas :

Can you confirm this is right please ? Or provide your formula for the beta volatility calculation please?

Cheers,
Ivan

[david-barnett]

Hi Ivan,

dist_calc_seq is a wrapper around dist_calc, which calculates distances between all samples. dist_calc_seq aims to make it convenient for the user to get the distance from the previous sample in a series into the sample data slot. The distance or dissimilarity you use is up to you to name.

If you set the dist argument as jaccard, you will get a jaccard distance. microViz calls phyloseq::distance, which calls vegan::vegdist functions to calculate the distance. However, the vegan jaccard distance is not the typically expected binary jaccard by default. https://search.r-project.org/CRAN/refmans/vegan/html/vegdist.html

dist_calc asks you to set binary = TRUE, or to tax_transform to binary first if you want a binary jaccard.

I notice now that the dist_calc_seq doesn't allow the former, which I may fix in a future version. For now, you can tax_transform to binary first if you want binary jaccard.

hope this helps,
David