SADI: Sequence too long

[Qi Zhou]

I am using SADI to do OMA. My data have 4 obervations with 3313 varialbes.
When I do: oma V1-V3313, subsmat(mvdanes) pwd(omd) length(3000) indel(1.5)
The error is One or both sequences too long: 3000 or 3000 > 1024
Also, when I do: oma V1-V3313, subsmat(mvdanes) pwd(omd) length(1024) indel(1.5)
The STATA 17 stopped and crashed.
What should I do to avoid this?
Many thanks!

[Jared Greathouse]

What is Sadi and why would you want to use it? Give me a small example of your data with dataex and the exact syntax you used with code delimiters.

[Brendan Halpin]

SADI is limited to sequences <= 1024 elements. While memory constraints are less than when SADI was written, the workspace required is proportional to the square of the sequence length, so the practical limit will rise slowly with memory growth. A related constraint is that every pairwise comparison will take a time proportional to the sequence length, so long sequences will make SADI slow.

I'd be interested to know what sort of sequences you are working with, and whether they really benefit from being so long. Maybe you could sample them down to a lower frequency (e.g., if it's monthly data, convert to years using the modal state in the year).