
Re: Principal Component Analysis Alternatives for low sample to dimensions ratio
Posted:
Apr 11, 2013 3:56 PM


yes Q factor analysis is one thing that can be done with a 2000 variable 300 case data matrix.
Yes Q factor analysis is an old time method of cluster analysis.
Depending on substantive nature of the data it may also be possible to cluster variables.
If the 2000 variables are word frequencies multidimensaional scaling is a possibility.
Many many thing are possible. But without the meaning of the data, level of measure, design role, etc. etc. all I can do is speculate.
What is the term David uses eesp?
Art Kendall Social Research Consultants
On 4/11/2013 2:36 PM, Rich Ulrich wrote: > On Wed, 10 Apr 2013 20:44:39 0700 (PDT), Krevin <kbrownk@gmail.com> > wrote: > >> Anyone know of good alternative methods to PCA when you have too many dimensions compared to samples? >> >> If I have 2000 variables and 300 samples, I cannot properly use PCA. >> >> I'm looking for something that can minimize false positive separation of sample points without needing to reduce my number of variables. > > As Art implies, it is usual for there to be some structure when there > are as many as 2000 variables. The sort of structure is apt to > matter for a constructive solution, assuming there is structure. > > I don't know what you have in mind when you say, "minimize > false separation of sample points..." but if you flip the matrix, > you have another conventional factoring (of samples). I think > of that as a sort of cluster analysis. Anyway, the samples > with low communalities will be ones that are relatively > "separate" from the others. >

