Dear Anatoliy,

   DA will not work on 2 columns by 4 rows. This is not nearly enough for the algorithm. 

Also, when you say "named as groups" do you mean that you toggled the variable state or do you literally mean you named them groups (also an error).

PAST is expecting something like

Group     Var1 Var2 ....

where having any less than 8 examples per group is just asking for trouble. This is not PAST's fault, just that you can't estimate variance with less and it is recommended to have much much more.

I have no idea. sorry. I don't use PAST.

Dear Bruce,

Thank you for your answer!

When I said "named as groups", I meant that the type of columns is "Group" (see attached file1).

I understand from your answer that I must add at least 8 variations per each parameter (for example as in the "file1" there are parameter1 "height" and parameter2 "fruits"). Am I right?

Best regards, Anatoliy

Dear Anatoliy,

   DA is intended to find the best equations to separate between two or more groups. You need to have adequate variation within each group before this process can begin. Second, when you label something like "Height" as a group variable, implicit is that you you would be interested is separating between every single possible height as though you were expecting them to be different in some multivariate space.

   I am pretty sure that you don't understand DA. Don't panic! But what question are you actually trying to answer and how are your data collected? Perhaps someone can help guide you to a more appropriate technique.

Dear Bruce,

Thank you for advices. I think I understood the problem. I.e. I must select one main GROUP and one or more SUBGROUPS in columns. For example (as in file1) it is name of species and as subgroups - age of individuals. And other columns should contain parameters which I have measured: height of indifiduals, number of flowers, width and length of leaves.

As a result, I got a graph "Orch-Cyp-Epi". Say me please, may I consider, that I understood the algorithm?

Best regards, Anatoliy

Yes, this is much more like it. DA will now tell you what equations best separate Orch v Cyp v Epi. Be careful about considering something like age as a grouping variable unless you will put it in a small number of categories (e.g. young/old)

Good luck.

Thank you for advice!

Dear Anatoliy and Bruce,

I have a simile type of DA to compare 7 rice varieties for 10 amino acids (6 replicates/sampling dates) in two different  locations. Can I use PAST 3 software to DA and how to enter the data into the software? Please advice me.

Hey all, I just wanted to join this topic and ask a question?

RE A CAP Test.

Will it work if I only have a factor that contains two groups and analyse 3 variables.

Basically I have used genetics to obtain two distinct clades and then I'm trying to run a test to see what significant characters are useful for discrimination of each clade.

Thank you in advance

Dear David

You really don't need CAP for this - it is not a multivariate problem. You can do it using ordinary linear regression, or logistic or other (depending on the assumptions you want to make about the separability of these clades).

Hi, I would like to ask if there is a way to test the statistical significance of the two main LDA axes and/or if there is something like posthoc tests for the group separation in PAST program

I would be happy if you can share with me some published studies that used LDA analysis

Thak you

Matteo Francioni, if I understood your question correctly you could use multivariate Hotelling's t-test, implemented in version 2 of PAST. You will compare by pairs of groups the differences between the (multivariate) means of different populations or groups. The t value can be associated to a p value.