One limitation to ICA approaches is that they cannot determine the number of appropriate components to decompose the data or their ordering. The number of components chosen is important as one too many may result in incorporating noisy sources and too few may mix the signals together. To determine the appropriate number of components, we used two different stopping methods applied to the eigenspectrum derived from PCA (13).The first stopping method is a measure of separability of the eigenvalues (λi), where we used a standard rule of thumb (36) to assess which eigenvalues (λi) exceed that expected from a random process. If the separation between eigenvalues (Δλi = λi − λi−1, for i > 1) falls below the uncertainty ∂λi = λi(2/n)2, then the component becomes more difficult to separate from its neighbor and from noise. The second approach uses Horn’s parallel analysis (37), a Monte Carlo simulation approach that randomly scrambles the data to create a suite of random samples and its simulated eigenspectrum and their uncertainties. If the observed λi exceeds the 95% confidence interval of the simulated λi, then the component is retained. We found that both methods indicate that four PCs are significant, and we used this number to decompose the data.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.