2
$\begingroup$

Using pdist() in the PST package, two probabilistic suffix trees (PSTs) can be compared to each other. The function will output a value for the degree of divergence between two fitted PSTs. However, how can this value be tested for statistical significance, to see whether it comes from a population where the value is greater than 0?

library(TraMineR)
library(PST)

a <- seqdef("A-A-A-B-B-C")
b <- seqdef("C-C-C-B-B-A")

a_pst <- pstree(a, ymin = 0.001, lik = FALSE, with.missing = FALSE)
b_pst <- pstree(b, ymin = 0.001, lik = FALSE, with.missing = FALSE)

pdist(a_pst, b_pst, l = 10, ns = 1000, symetric = TRUE, output = "mean")

Would it be sufficient to do:

output <- pdist(a_pst, b_pst, l = 10, ns = 1000, symetric = TRUE, output = "all")
t.test(output)

?

$\endgroup$
2
  • 2
    $\begingroup$ you have introduced new tags pst and vlmc. Can you please add tag wikis? $\endgroup$ Aug 6, 2017 at 19:46
  • $\begingroup$ @kjetilbhalvorsen done--we can delete these comments now. $\endgroup$
    – histelheim
    Aug 9, 2017 at 17:26

1 Answer 1

0
$\begingroup$

A t-test is probably not appropriate because the divergence can only be positive.

You may find a statistical test for the divergence between two PST models in Christine Largeron (2003).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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