[R-sig-eco] Adonis for significance of clusteredness from hclust (vegan package)

[Pierre THIRIET]

Dear Ansley,

I agree with Zoltan.

I suggest SIMPROF test in package "clustsig". A complimentary package 
for use with hclust; simprof tests to see which (if any) clusters are 
statistically different. The null hypothesis is that there is no a 
priori group structure.
See Clarke, K.R., Somerfield, P.J., and Gorley R.N. 2008. Testing of 
null hypothesis in exploratory community analyses: similarity profiles 
and biota-environment linkage. J. Exp. Mar. Biol. Ecol. 366, 56-69

HTH
Pierre

Le 04/10/2016 à 14:14, Zoltan Botta-Dukat a écrit :
> Dear Ansley,
>
> I cannot answer your question, I hope someone else will answer. I'd
> rather point out a problem in your approach. Statistical tests were
> developed for testing difference between a priori groups, thus estimated
> Type I error rate is valid only for this situation. When you calculates
> Type I error rate for comparison of groups created by cluster analysis
> of the SAME data, the calculated error rate will be lower than the valid
> error rate. So you cannot use the term "significant" in this situation.
>
> Sorry for making you sadden by this information.
>
> Zoltan
>
> 2016.10.03. 21:52 keltez�ssel, Ansley Silva �rta:
>> Hello:
>>
>> I have created a dendrograms using hierarchical cluster analysis with the
>> vegan package (function: hclust).
>>
>> By visually observing the dendrogram, I have determined that there are 3
>> main clusters if I "cut" the tree at the height 0.25  (please see the
>> dendrogram from the code).
>> I then created a new dataset, which is essentially the same as the
>> original, but I have added the categorical variable Group to represent
>> these 3 main clusters.
>> ST0 is group a, AP0 and AP100 is group b, and AP200 AP300 ST100 ST200 ST
>> 300 is group c.
>> I want to now if they are significantly different from each other.  I
>> understand, from the output pasted below, that I can accept that there is a
>> significant effect of Group.  Is this the only thing I can say from
>> Permanova?  What would be the code for a follow up test to look at
>> pair-wise significant differences?
>> Thanks very much.
>>
>> Call:
>> adonis(formula = species ~ Group, data = environ, permutations = 999)
>>
>> Permutation: free
>> Number of permutations: 999
>>
>> Terms added sequentially (first to last)
>>
>>             Df SumsOfSqs  MeanSqs F.Model      R2 Pr(>F)
>> Group      2   0.40244 0.201219   4.969 0.66528  0.007 **
>> Residuals  5   0.20248 0.040495         0.33472
>> Total      7   0.60492                  1.00000
>> ---
>> Signif. codes:  0 �***� 0.001 �**� 0.01 �*� 0.05 �.� 0.1 � � 1
>>
>>
>>
>>
>> _______________________________________________
>> R-sig-ecology mailing list
>> R-sig-ecology at r-project.org
>> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
>
>
>
> _______________________________________________
> R-sig-ecology mailing list
> R-sig-ecology at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-ecology


	[[alternative HTML version deleted]]