EViews.com

I am running PCA on 35 different variables including gdp, inflation, consumption and etc. After doing PCA on those variable, could anybody tell me how I could seperately find the explainibility of first component only on gdp, consumption and inflation? To make finer detail, after gaining the result below, how can I use only first components (first eigenvalues) for only selected variables by using eviews?



Principal Components Analysis
Date: 07/18/14 Time: 08:10
Sample (adjusted): 2000M03 2014M03
Included observations: 57 after adjustments
Balanced sample (listwise missing value deletion)
Computed using: Ordinary correlations
Extracting 37 of 37 possible components


Eigenvalues: (Sum = 37, Average = 1)
Cumulative Cumulative
Number Value Difference Proportion Value Proportion


1 22.44237 18.39826 0.6066 22.44237 0.6066
2 4.044106 1.348658 0.1093 26.48647 0.7159
3 2.695448 0.483347 0.0728 29.18192 0.7887
4 2.212101 0.738559 0.0598 31.39402 0.8485
5 1.473542 0.558126 0.0398 32.86756 0.8883
6 0.915416 0.149788 0.0247 33.78298 0.9131
7 0.765627 0.228788 0.0207 34.54861 0.9337
8 0.536839 0.086991 0.0145 35.08545 0.9483
9 0.449848 0.137295 0.0122 35.53529 0.9604
10 0.312554 0.027246 0.0084 35.84785 0.9689
11 0.285308 0.083460 0.0077 36.13316 0.9766
12 0.201848 0.064610 0.0055 36.33500 0.9820
13 0.137239 0.005498 0.0037 36.47224 0.9857
14 0.131741 0.029419 0.0036 36.60398 0.9893
15 0.102322 0.028505 0.0028 36.70631 0.9921
16 0.073817 0.020942 0.0020 36.78012 0.9941
17 0.052875 0.006430 0.0014 36.83300 0.9955
18 0.046445 0.014492 0.0013 36.87944 0.9967
19 0.031953 0.014069 0.0009 36.91140 0.9976
20 0.017884 0.002827 0.0005 36.92928 0.9981
21 0.015057 0.004187 0.0004 36.94434 0.9985
22 0.010870 0.001089 0.0003 36.95521 0.9988
23 0.009781 0.002108 0.0003 36.96499 0.9991
24 0.007673 0.000686 0.0002 36.97266 0.9993
25 0.006988 0.000937 0.0002 36.97965 0.9994
26 0.006051 0.002191 0.0002 36.98570 0.9996
27 0.003860 0.001073 0.0001 36.98956 0.9997
28 0.002787 0.000457 0.0001 36.99235 0.9998
29 0.002330 0.000612 0.0001 36.99468 0.9999
30 0.001718 0.000446 0.0000 36.99640 0.9999
31 0.001273 0.000337 0.0000 36.99767 0.9999
32 0.000935 0.000206 0.0000 36.99860 1.0000
33 0.000730 0.000481 0.0000 36.99933 1.0000
34 0.000249 6.30E-05 0.0000 36.99958 1.0000
35 0.000186 4.97E-05 0.0000 36.99977 1.0000
36 0.000136 3.99E-05 0.0000 36.99990 1.0000
37 9.61E-05 --- 0.0000 37.00000 1.0000

I'm not sure that I understand the question.

Let me make it simpler. In 35 different variables, there are gdp, consumption and inflation. Then, by using one componet generated from PCA which is the highest eigenvalue, do you know how to associate this first componemt only with three variables, gdp, consumption and inflation?

I'm not certain that calculation makes sense since the PCA computation is for the full set of 35 variables and the ordering of the eigenvalues is with respect to the full variance or correlation matrix. I don't believe it is possible from the PCA results. I suppose that you could construct component scores for gdp, consumption, and inflation using only the first component, but after that it's not clear how one would relate the variance associated with those scores with the variance of the original three variables.

Having said that, there may be some deep decomposition result that I'm missing that will allow you to do what you wish. Perhaps someone else who works with these models will be able to help.

hello guys
I am trying to create an Index with 4 variables using the principal component analysis. I am not familiar with how this works :( please any form of assistance would be greatly appreciated.
thanks.

I usually hesitate to tell people to read the manual, but in this case, I don't think there's anything I would add to what we wrote there and this question has gone unanswered for a bit.

Hi, is it OK to make the principal components using levels instead of rate of change? (this is for creating and index with stocks, I know in the manual the example is done with weekly rate of changes)