Reduce data dimension using PCA

pca() outputs the coefficient of the variables and principal components of a data. Is there any way to reduce the dimension of the data (340 observations), let say from 1200 dimension to 30 dimension using pca()?

[Adam]

You should just be able to keep the 30 largest components from running pca.

[Hg]

I use

[residuals,reconstructed] = pcares(X,ndim)

Answer 1

[Vassilis Papanastasiou]

Hi Hg,

What you can do is to use pca directly. Say that X is of size 340x1200 (340 measurements and 1200 variables/dimensions). You want to get an output with reduced dimensionaty of 30. The code below will do that for you:

p = 30;

[~, pca_scores, ~, ~, var_explained] = pca(X, 'NumComponents', p);

• pca_scores is your reduced dimension data.
• var_explained contains the respective variances of each component.

I hope that helps.