NMDS question

[Austin M]

Hi all,

My name is Austin Menzmer and I am a graduate student doing some community ecology data analysis in Georgia. My PI recommended I purchase PC-ORD both for its' capabilities to do MRPP and NMDS. I purchased both the program + Dr. Peck's Multivariate Analysis for Ecologists over the past few weeks (I also have Analysis of Ecological Communities). I'm happy to report that running MRPP worked just fine; however, I'm having some issues getting the NMDS to work on my data. 

My first course of action was to follow along with a tutorial I found on YouTube (https://www.youtube.com/watch?v=gXFUy3zNQmU), using the same (Oak Woods Simple) dataset. I was easily able to import the Species sheet as the main matrix, and SecondMatrix sheet as the second matrix. I then ran an NMS through the Ordination menu, selected Sorensen as my distance measure, and chose medium autopilot mode.  Once the computing was done, I clicked Graph -> Graph Ordination -> 2D -> Joint plot/biplot to get a first draft of an NMDS graph/figure. Everything worked as it did in the video.

Now, my dataset has vegetative but no environmental data. So while I have percent cover data with species as columns and sampling units as rows, I don't have any other data outside of my three different treatments (my experiment focusing on effects of disturbance, so I have a control treatment, low severity treatment, and high severity treatment). 

I noticed on the Oak Woods example dataset that it appeared that Landform was a grouping variable for which treatments the data was categorized into, so on the OakWoodsSimple dataset I created a third sheet called "2ndMatrix" where I just kept the Stand and Landform columns. Importing "2ndMatrix" as my new second matrix, I repeated the NMS as described before (with the exception of classifying Landform as categorical). Sure enough, when I went to graph the NMDS, joint plot/biplot worked and produced a graph, only providing an informational warning that 2ndMatrix contained no quantitative data (okay, that's what I expected, anyways). 

So, my next step was to run the same analyses on my own data. My main matrix of vegetative data was in the same format as OakWoodsSimple, and my 2ndMatrix for my own data (site name WS) has Treatments (analogous to the Landform name of OakWoodsSimple) of 1, 2, or 3. I ran the NMDS with the same settings as I did with OakWoodsSimple. 

A few interesting developments:

1) My graph and columns outputs for both sampling units (A1, E7, etc.) and species (Vitis rotundifolia, Smilax glauca, etc.) have only one column of data each, rather than 2-3 like the OakWoodsSimple outputs had. Not sure what to make of that, how to explain, or how to fix. 

2) When I went to Graph -> Graph Ordination -> 2D, the Joint plot/biplot button is greyed out, despite me (to my knowledge) having with my data in the same format as the OakWoodsSimple dataset. Not sure how to get the program to produce a NMDS graphic. 

I'm honestly not sure how to proceed or what to do next, even after reading through pertinent chapters of both books. I've attached a screenshot of my data in PC-ORD.  Would anyone be willing to give an excited but confused PC-ORD newbie tips on how to get an NMDS to run and graph successfully? Any tips or nudges in the right direction would be much appreciated.

Thank you ever so much!

Austin Menzmer

(alternate contact): austin....@uga.edu

[Bruce McCune]

Austin, it sounds like NMS is thinking that a 1D solution is best for your data, hence only 1 col of graph scores and no 2D plot. This might or might not be a problem. If there is only one controlling factor, then it is fine. But you may have a need to transform your data or deal with an outliers. Plenty to read on these topics in the PC-ORD help system and your books!

Bruce McCune

--
You received this message because you are subscribed to the Google Groups "PC-ORD" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pc-ord+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pc-ord/0a033989-83c3-4339-8219-782428a7bac6n%40googlegroups.com.

[Austin M]

Dr. McCune,

Ahhh, that makes sense. Yes, the NMS output did say the a 1D solution is best for my data. I will look into how to transform data in the other resources available. 

Thank you so much for your time!

Austin