How to exclude grid-cell with just 1 or few species (label) from a Cluster Analysis

17 views
Skip to first unread message

marco iocchi

unread,
Feb 3, 2021, 3:41:24 PMFeb 3
to Biodiverse Users
Hello everyone,
I want to perform a Cluster Analysis only for those grid-cells that contain more than x species (label), excluding those cells that contain just few species (less than x).
Is it possible to set this constrain?
I'd like to compare the resulting dendrograms obtained setting different thresholds of this constrain (x).

Cheers

Marco

marco iocchi

unread,
Feb 3, 2021, 5:07:52 PMFeb 3
to Biodiverse Users
Hello everyone,

I found the solution by myself
I used the "Run Exclusions" option of the "BaseData" Menu

Cheers

Shawn Laffan

unread,
Feb 5, 2021, 5:51:26 PMFeb 5
to biodiver...@googlegroups.com, marco iocchi
Hello Marco,

Good to hear you worked it out. 

The issue with removing groups is that you can reduce the range and abundance estimates for some of your labels, which could then be an issue for endemism and rarity calculations.

The approach in that case is to run a spatial analysis to calculate richness, and then use a definition query in the cluster analysis to only consider cells (groups) with more than your richness threshold. 

The steps are:

1. 
Run a spatial analysis with a single neighbour set of sp_self_only().  This is the default neighbour set in v3, so you should not need to do anything except calculate the richness indices.  The next step assumes this is called "sp1".

2.
Run the cluster analysis, specifying the definition query to be like the one below.  Edit the "> 3" at the end to change the threshold.
sp_get_spatial_output_list_value (
    output  => 'sp1',              #  using spatial output called sp1
    list    => 'SPATIAL_RESULTS',  #  from the SPATIAL_RESULTS list
    index   => 'RICHNESS_SET1',    #  get index value for RICHNESS_SET1
) > 3                              #  and return true if the value is >3


That condition is probably more than most people would want to type (or copy and edit), so I have opened an issue to implement a simpler approach in version 4.
https://github.com/shawnlaffan/biodiverse/issues/783

The approach above does have some advantages, though.  The main one is that one can apply any amount of complexity in the neighbour sets, and threshold using any index value.  For example, one could calculate phylogenetic diversity (PD) using bioregions as the neighbour sets, and then threshold on those values.  Then one would exclude any cell in a bioregion with PD less than the desired threshold. 


All that said, if your cells with few labels are likely to be errors then it is better to use the Run Exclusions approach.  And if you are not interested in rarity or endemism type indices then the Run Exclusions approach will work perfectly well.


Regards,
Shawn.




On 4/02/2021 9:07, marco iocchi wrote:
Hello everyone,

I found the solution by myself
I used the "Run Exclusions" option of the "BaseData" Menu

Cheers

On Wednesday, February 3, 2021 at 9:41:24 PM UTC+1 marco iocchi wrote:
Hello everyone,
I want to perform a Cluster Analysis only for those grid-cells that contain more than x species (label), excluding those cells that contain just few species (less than x).
Is it possible to set this constrain?
I'd like to compare the resulting dendrograms obtained setting different thresholds of this constrain (x).

Cheers

Marco
--
You received this message because you are subscribed to the Google Groups "Biodiverse Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biodiverse-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biodiverse-users/e65cb201-515a-436e-a392-608c5c9e64a4n%40googlegroups.com.

marco iocchi

unread,
Feb 6, 2021, 4:28:47 AMFeb 6
to Shawn Laffan, biodiver...@googlegroups.com
Hello Shawn,

I wasn't interested in endemism and rarity, but your suggestion is very helpful and inspiring because it leads me to explore further analysis on my data

thank you
--
Marco Iocchi

Dipartimento di Scienze
Università degli Studi di "Roma Tre"
Viale Marconi, 446 - 00146 - Roma

Shawn Laffan

unread,
Feb 7, 2021, 3:06:51 AMFeb 7
to biodiver...@googlegroups.com, marco iocchi
You're welcome Marco. 

Please do ask if you have more queries. 

Regards,
Shawn.
Reply all
Reply to author
Forward
0 new messages