Carlos Guerra
2006-Nov-16 11:08 UTC
[R] Problems with principal components analysis PCA with prcomp
Dear friends,
I am beginning to use R software in my academic research and I'm having some
problems regarding the use of PCA.
I have a table with 24445 rows and 9 columns, and I used the function
prcomp() to do the analysis.
Working with an example?:
x<-read.table("test.txt", header=T)
row.names(x)<-x[,1]
x<-x[,-1]
require(stats)
pca<-prcomp(x, scale=T)
names(pca)
## here begin my problems? know what???##
My questions are:
1. How can I group my samples (rows) in a graphical way?
2. How can I extract the samples related to each group? ?or? How can I
create a new column with "n" values corresponding to each group
created?
3. Is it possible to extract/export the information in an object into a file
with a dbf extension (e.g. test.dbf)?
Best regards,
Carlos
+++++++++++++++++++++++++++++++
Carlos GUERRA
Instituto Polit?cnico de Viana do Castelo
Escola Superior Agr?ria de Ponte de Lima
Mosteiro de Ref?ios do Lima
4990-706 Ponte de Lima
PORTUGAL
MAIL: carlosguerra at esa.ipvc.pt
TEL: +351 912 407 109
+++++++++++++++++++++++++++++++
-------------- next part --------------
FID ALTIM TEMP CLI FAS
0 14 4 1 2
1 65 1 1 3
2 54 1 7 1
3 112 1 5 1
4 117 4 2 1
5 180 2 1 2
6 54 4 1 3
7 125 4 1 2
8 217 4 2 2
9 171 3 2 2
10 37 4 3 3
11 114 2 4 3
12 229 4 4 3
13 286 3 3 3
14 227 3 5 3
15 16 4 5 2
16 74 4 5 2
17 89 4 6 1
18 98 2 5 1
19 64 2 4 2
20 96 2 6 1
21 100 4 6 3
22 147 4 7 1
23 196 4 7 3
24 282 3 1 3
25 308 3 1 3
Sarah Goslee
2006-Nov-16 14:09 UTC
[R] Problems with principal components analysis PCA with prcomp
Hello Carlos, # let me save you a couple steps here> x<-read.table("test.txt", header=T, row.names=1)> pca<-prcomp(x, scale=T) > names(pca) > > ## here begin my problems? know what???##Try this for a graphical view of your data: biplot(pca) You might also want to work through the examples given in ?prcomp and ?princomp.> My questions are: > 1. How can I group my samples (rows) in a graphical way? > 2. How can I extract the samples related to each group? ?or? How can I > create a new column with "n" values corresponding to each group created?Here I see a problem... if you are interested in _groups_ in the data, you might be better off with a clustering method, of which R has an abundance. Ordination is generally about _gradients_ rather than groups - continuous change vs discrete groups. If you are determined to divide your ordination into groups, one possibility is to use conditional subsetting on the ordination values.> 3. Is it possible to extract/export the information in an object into a file > with a dbf extension (e.g. test.dbf)?help.search(dbf) finds the function write.dbf - sounds suspiciously useful. Sarah -- Sarah Goslee http://www.functionaldiversity.org