Hi there,
I am just wondering whether correction for multiple testing is necessary with the indicator value IndVal by Dufrene & Legendre (1997), as well as the modified version by De Caceres et al. (2010). This correction is sometime proposed (e.g. by Legendre 2009), but not implemented in the R code of packages 'labdsv' and 'indicspecies' where the issue is at least mentioned.
As I understood, one major advantage of the IndVal over TwinSpan analysis was that is was not affected by (in this case the abundances of) other species.
This advantage of unaffectedness would be lost if correction for multiple testing would be conduncted for the IndVal, since the total number of species (hence number of tests) would then affect the possibility for a single species to gain significance with respect to its indicator value.
Any suggestions?
Thanks!
Armin
Hi Armin,
I am not sure but just guessing here: as the IndVal tests each species separately for the significance of its indicator value, i.e. each species is a separate test where the response changes, shouldn´t the tests then be considered independent and thus no correction would be required?
again, I am just blowing up some thoughts...
greetz
Hi Jens,
thanks for your answer! I had similar thoughts and would intuitively agree with you. I was just unsure, since Pierre Legendre who co-developed this index recommended correction for multiple testing.
Cheers
Armin
Hi Kriss,
I did not find a final answer to the question. Both ways have valid arguments. And I didn't send an email to Pierre Legendre, to be honest.
Armin
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology