Fit models for the different number of components
Perhaps decrease the maximal number of iterations to 500 if you are in a hurry, but don’t do that in “real life”. For each fitted model, the model of the data being Mf
(f the number of components), can be obtained using the file nmodel
.
Calculate residuals as
Ef = X - Mf;
Use these residuals for plotting purposes as was done with the raw data in the introductory chapter. Use the residuals for calculating the sum-squared residuals
(or obtain these directly in the output from the algorithm) for scree-plots.
For calculating the core consistency use the m-file corcondia
.
For the model with loadings in the vector Factorf
calculate the core consistency ascf = corcond(X,Factorf);
Where f is the number of components.
Compare the core consistencies for various numbers of components:
plot([1:5],[c1 c2 c3 c4 c5])
xlabel('Number of components')
ylabel('Core consistency')