2
$\begingroup$

I am trying to compute a distance matrix using the method "SVRspell" (seqdistOO function).

I want to use an exponential weight of spell length equal to 2 and I want to discard matching sub-sequences of length one. No states' similarity matrix.

Note: sequences in the 'seqs' object below have a length of 25 (ages 15 to 39) and the alphabet has 20 states (marital status plus children ever born).

So, I run this

>svr<-seqdistOO(seqs, method="SVRspell", tpow=2, kweights=sqrt(c(0,2:25))) [>] 6315 sequences with 20 distinct events/states [>] 5023 distinct sequences [>] min/max sequence length: 1/11 [>] computing distances using NMSMST metric [>] total time: 30.41 secs

I do not understand why it seems to be using a NMSMST metric instead of the SVRspell with the specified parameters. In my understanding SVRspell is not equivalent to NMSMST

Am I coding something wrong? or is it because SVRspell pertains to the 'family' of NMS metrics?

Thanks,

Andres

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ You might be better on a site specific to your software unless you want to separate out the statistical part from the programming part. $\endgroup$
    – mdewey
    Commented Mar 7, 2017 at 18:30

1 Answer 1

Reset to default
1
$\begingroup$

Your code is correct and the "... using NMSMST" message is just a display bug.

Actually, the NMS, NMSMST, and SVRspell methods proposed by the package seqdist2, all use a same algorithm. The difference is controlled by a few internal parameters.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you, it is good to know I was doing the right thing. $\endgroup$
    – Andres Felipe Castro Torres
    Commented Aug 24, 2017 at 7:57
  • 1
    $\begingroup$ Package seqdist2 is now obsolete. Its content (with many improvements) has been made directly available in TraMineR since v 2.0-5. See the documentation page of the seqdist and seqcost functions. $\endgroup$
    – Gilbert
    Commented Aug 24, 2017 at 11:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.