JMP > Factor Analysis > Input Structure Issue

[#1]

Hi JMP Community,

I have a large data set (354 cases [Rows] x 4800 variables [Columns]) derived from a metabolomic analysis. While reviewing the data by PCA, I noticed a clear separation of the data into two populations (see below). Now, I need to assess the possible contribution of cases' properties (e.g., Gender, Age, Treatment) to this separation.

I assumed that Factor Analysis would be the right platform to use, but I failed to grasp how to format the input for this analysis properly.

I cannot share the actual data because of its confidential nature. Hence, let me know if it would help if I created and shared a mock table.

Best,

TS

[#2]

Hi @Thierry_S , did you check the two examples in the scripting index?

Perhaps that may guide you through. Good Luck.

Names Default To Here( 1 );
dt = Open( "$SAMPLE_DATA/Cities.jmp" );
obj = dt << Principal Components(
	Y( :OZONE, :CO, :SO2, :NO, :PM10 ),
	"on Correlations"
);
obj << Factor Analysis( "ML", "SMC", 2, "Varimax" );

[#3]

Hi @Georg ,

Thank you for reminding me of the examples and the scripting guide. I did study these examples, but I still struggle with applying this method to my large data set. Does it even make sense to look at 4800 variables in Factor Analysis? I am starting to believe that my original question was misguided.

Best,

TS  

[#4]

No problem @Thierry_S ,

I do not have experience with that many factors.

Maybe it's a good idea to have an example table, to get further ideas from the community. 

As you have already identified the clusters (i.e. the "y"), what about response screening for identifying (and filtering) important factors? I believe, you have to test different ways and compare.