Running Permanova in Primer on large datasets

#1
Hi all

I am new to Talk Stats. Have a quick question regarding computation times to run permanova in Primer. I am trying to gauge how long it would take to run a univariate analysis on a dataset with resemblage matrix of approx 5000 rows/cols, for a nested design with 5 factors permutations set to 9999. On my current PC - I started the analysis and 12 hrs later, it appears to be about 10% done. Would like to know what hardware other Primer users have and how long similar analyses have taken? (thinking I might need to invest in a new setup!).

Thanks
 

bugman

Super Moderator
#2
Well, I use an i7 with 8GB of ram and for that size data set I would expect anywhere between 24 and 36hr.

I often leave these analyses running over night at work (8-10 hrs) and that still gives me time for a quick coffee in the morning before its done, but I usually have <300 rows/columns and use 999 runs.

Why not reduce your permuations to 999/ With such a large dataset, this should be more than adequate.


Oh and P.S. Welcome to talkstats!
 
#3
Thanks bugman

I agree - 999 should be fine, but I have a client who is specifying 9999 (to be consistent with other components of the project). The dataset I have running is a dummy one, set up to resemble how large the monitoring project will end up in another 1-2 yrs. I've got an i3 and also 8G RAM. The analysis has been going about 39 hrs now, and progress bars looks at about 40% (need to set it going before heading away for a weekend! ;)). I've been told about a processor that is 'multi thread, multi core' with a speed of 3.8GHz - thinking if I coupled that with double the RAM it might be able to get it done faster. Have to see if it would be worth the $$ - but was curious what speeds other users have found and what their setups are.

cheers