pca with more than one grouping variable

I want to do a pca plot of 20 samples I measured. For this I used gscatter. However, I want

1) to color my samples with 7 colors for the different type of organisms i measured and in addition

2) I want to give symbols for another variable called, dead or alive where I want to use square for dead and triangles for alive

How can I create a pca plot showing what I described above? do I need to use something else than gscatter?

Answer 1

[Jeff Miller]

gscatter will accept multiple grouping variables, in your case organism & dead/alive. There is an example in the gscatter documentation--search for "Scatter Plot with Multiple Grouping Variables". I believe you specify 14 colors and 14 symbols for your 7*2=14 combinations of the two grouping variables.

[Kim Arnold]

dear jeff,

thanks for your answer. yes I already tried this but what i want is that the 7 colors define the types of organisms and they should stay the same, the variable dead/alive should just make them squares or triangles but they should keep the color they received by the first grouping. The example they give in the gscatter would be possible but i dont want to use 14 colors and so many symbols, for the viewer this looks very complicates. However, thanks for your input.

Best,

Kim

[Jeff Miller]

Hi Kim,

Sorry I wasn't clearer about my suggestion. I had in mind that you would specify 14 colors and 14 symbols with a structure that reflected your groups, something like this:

gscatter(x,y,g,'rrkkggbbccmmyy','o*o*o*o*o*o*o*',8,'on','Age','Weight')

I think this would show the first two groups (i.e. organism 1 alive & dead) as red, with different symbols, the next 2 groups (i.e., organism 2) as black with the same symbols distinguishing alive/dead, etc.

Jeff

[Kim Arnold]

Dear jeff,

no worries, you were clear enough. I tried already what you suggested but this did not work correctly.

I finally figured out how it works. If you have like 17 samples and you have 6 groups (my mistake i wrote 7 earlier, but does not really matter regarding this aspect) of organisms which you want to color with 6 colors and then give in addition the symbols dead o and alive * you need first to know if you have in each of the 6 groups dead and alive or only dead or only alive because this makes a difference for the code and is something i did not mention to you when I described my problem. What I did first is the labels of the grouping and then I created

a logical array where 0= dead and 1=alive as [0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0].

Then I merged those two grouping variables as described in the example. Next you define

6 colors for your groups as 'rkkgbcmy'

Then this was the part which was first not clear to me. I went through my 6 groups. first group I had only dead --> symbol o, 2nd group I had dead and alive --> symbols o*, 3rd group only dead--> symbol o, 4th group only alive--> symbol *, 5th group only dead --> symbol o and 6th only dead --> symbol o which gives you an entry for 7 symbols (not 6!) because only in one group you have dead alive and need to specify 2 symbols.

'oo*o*oo'

However, it took a while to figure out, thanks for your input :)

Best,

Kim

[Jeff Miller]

Hi Kim,

You are welcome. That does sound pretty complicated, so extra-good that you got it working.