# NMDS ordination and stress value

Dear all

I am a newbie and I apologize in advance for any ignorant/silly questions I might ask

I am working on ant gut microbiota and I have about 120 samples. Those 120 samples are divided into 2 groups: A and B. I have calculated the braycurtis distances and tried to plot them in 2 axes, however when I do that the lowest stress value is above 0.33 and the R-squared value is 0.47… even if I run the nmds with 3 dimensions the values dont improve much, he lowest stress value becomes 0.22 and the R-squared 0.62
I have to use 4 dimensions and then the lowest stress value becomes 0.167349 and the R-squared 0.717444!
As far as I know since I use braycurtis, then I have to use NMDS ordination, correct? so what should I do? Consider the above values acceptable?

And I also have a second question: I can see that if I plot the samples from group A and group B, while almost all samples from group B pretty much cluster in the same area, the samples from group A are spread in every direction. I was wondering if there are any other ways to show/prove that samples from group A are varying more (more spread) than samples from group B, besides plotting it?
Any ideas are welcome!

As far as I know since I use braycurtis, then I have to use NMDS ordination, correct? so what should I do? Consider the above values acceptable?

This is common when you have a large number of samples. You can also use PCoA. Remember, that these ordination techniques are data visualization tools. They aren’t a statistical analysis. If you see patterns in the ordination, you’ll have to do further work to show that the patterns are meaningful and not just seeing Mickey Mouse in the clouds

And I also have a second question: I can see that if I plot the samples from group A and group B, while almost all samples from group B pretty much cluster in the same area, the samples from group A are spread in every direction. I was wondering if there are any other ways to show/prove that samples from group A are varying more (more spread) than samples from group B, besides plotting it?

You’ll want to check out the homova command in mothur.

Good luck!
Pat

I absolutely agree with you that the ordination is only for visualization, I was only wondering if there is a way to find the “perfect” or the “best possible” ordination method for visualization, I will also check how it looks with a pcoa even if I would rather keep the NMDS!

Yes I checked the HOMOVA command, its perfect! The variation within group is exactly what I needed, so I already got all the values for the variation within each of my groups and I am “playing” with them a bit more

Thanks