PCA Output?

This may all be correct, but in the end psych delivers a well interpretable output and that should be the essence from the user’s point of view.

The core of my solution (for learning purposes) is this (PCA_Data contains the observations, PCA_HK the estimated number of main components):

PCA_Modell = R"principal($PCA_Daten, $PCA_HK)"
PCA_scores =  R"principal($PCA_Daten, $PCA_HK)$scores"
show(PCA_Modell)

As you can see, I use R over RCall. And that is the output:

RObject{VecSxp}
Principal Components Analysis
Call: principal(r = `#JL`$PCA_Daten, nfactors = `#JL`$PCA_HK)
Standardized loadings (pattern matrix) based upon correlation matrix
     RC1   RC2   RC3   RC4   h2   u2 com
1   0.91 -0.13  0.07  0.02 0.85 0.15 1.1
2  -0.34 -0.02 -0.72 -0.13 0.64 0.36 1.5
3   0.75  0.04  0.44  0.23 0.81 0.19 1.9
4   0.35  0.46 -0.03 -0.25 0.39 0.61 2.5
5   0.13  0.00 -0.27  0.80 0.73 0.27 1.3
6  -0.14  0.88  0.03  0.07 0.80 0.20 1.1
7  -0.02  0.88 -0.11  0.09 0.79 0.21 1.1
8   0.78  0.10 -0.44  0.01 0.80 0.20 1.6
9  -0.73  0.01 -0.01 -0.25 0.60 0.40 1.2
10  0.15  0.06  0.27  0.76 0.68 0.32 1.4
11 -0.23 -0.11  0.78 -0.13 0.69 0.31 1.3

                       RC1  RC2  RC3  RC4
SS loadings           2.87 1.80 1.67 1.44
Proportion Var        0.26 0.16 0.15 0.13
Cumulative Var        0.26 0.42 0.58 0.71
Proportion Explained  0.37 0.23 0.21 0.19
Cumulative Proportion 0.37 0.60 0.81 1.00

Mean item complexity =  1.4
Test of the hypothesis that 4 components are sufficient.

The root mean square of the residuals (RMSR) is  0.09
 with the empirical chi square  1473.01  with prob <  3.2e-303

Fit based upon off diagonal values = 0.89

The output is easy to read and contains some information on the quality of model estimation. This is basic information for the data analyst.

The value of the latent variables (scores) is also supplied simply (well, Julia (or the function of the package) also supplies this). And by the way, the scree-graphic is delivered as well.

The effort I have to make as a user is quite small. Speaks from my point of view for the R-function. I still see potential for the MultivariateStats package.

Thank you for your support and greetings,
Günter

1 Like