Skip to content

Rearranging PCA results from R

2 messages · psycrcyo, Uwe Ligges

#
Hi!! 
I'm having trouble selecting 10 out of 41 attributes of the KDD data set. In
order to identify the components with the higher variance I'm using
princomp. the result i get for summary(pca1) is:


                              Comp.1            Comp.2          Comp.3             
Comp.4        Comp.5              Comp.6            Comp.7           Comp.8          
Comp.9           Comp.10
Standard deviation     9.882181e+05  3.303966e+04  7.083767e+02 
3.282215e+02  9.839173e+01 4.642758e+01  2.923245e+01  6.447245e+00 
2.689471e+00  1.292525e+00

Proportion of Variance 9.988828e-01  1.116555e-03  5.132601e-07 
1.101902e-07  9.902073e-09  2.204758e-09  8.740565e-10  4.251648e-11 
7.398482e-12  1.708784e-12

Cumulative Proportion  9.988828e-01 9.999994e-01 9.999999e-01 1.000000e+00
1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00 1.000000e+00
1.000000e+00

and for the loadings a constant 0.024 for the proportion of variability:

                    Comp.1 Comp.2 Comp.3 Comp.4 Comp.5 Comp.6 Comp.7 Comp.8
Comp.9 Comp.10 
SS loadings     1.000   1.000    1.000   1.000   1.000    1.000    1.000  
1.000   1.000   1.000   
Proportion Var  0.024    0.024    0.024  0.024    0.024   0.024    0.024  
0.024   0.024   0.024   
Cumulative Var  0.024    0.048    0.071  0.095    0.119  0.143    0.167  
0.190   0.214   0.238  

So the questions are: Which of the two is the right proportion of variance?
and, is there a way for R to tell me which attributes they belong to?

Any help will be very appreciated.

psycrcyo



--
View this message in context: http://r.789695.n4.nabble.com/Rearranging-PCA-results-from-R-tp3467015p3467015.html
Sent from the R help mailing list archive at Nabble.com.
#
On 22.04.2011 00:36, psycrcyo wrote:
Actually you calculated the first 10 principal components. You have not 
selected anything - particularly no "attributes", all "attributes" are 
included in your 10 first PCs. I'd suggest to read some textbook about PCA.

Some people like to perform stepwise regression of variables on the 
first PC if it explains a lot of the variance, but that should be done 
*very* carefully, if at all.

Best,
Uwe Ligges