Selecting the number of principal components: estimation of the true rank of a noisy matrix (Q682300): Difference between revisions

The class of statistical methods utilizing a hypotheses testing framework for determining the rank of a singular matrix in a noisy matrix model is proposed. The estimated rank corresponds to the number of principal components to retain in principle component analysis. Under Gaussian assumption, the proposed hypothesis testing method provides exact type I error controls along with exact confidence intervals of signal parameters in finite samples.