Child pages
  • PCA Tool - Images of signal vs sample
Skip to end of metadata
Go to start of metadata

The PCA tool runs principal components analysis on an image composed of spectra/diffraction patterns of multiple samples.

Example

This file contains a sequence of 10 IR absorption spectra ([10,1602] data block), shown here as separate lines. To use the PCA tool to analyse the variance in this data set:

  • Plot the data as an image,
  • From the tools menu, select PCA
  • From the PCA View, Select the number of PCs to calculate and click Run PCA.
  • When the calculation is complete, the % variance explained should be plotted against the component number in the Variance Explained tab
  • There are other tabs with different ways to explore the PCA results
  • The Scores/Loads tab allow the difference PC scores to plotted against each other (or the y-axis data of the image, which might be, for example, Temperature)
  • The plot on the left shows the average spectrum/pattern plotted with the loadings vector for the selected PCs
  • The Reconstruct tab allows the data to be reconstructed using a limited number of PCs. This can be used to see how many components are needed to explain the variance in each sample.
  • The number of PCs to use can be changed by adjusting the PCs spinner, the sample to investigate can be changed using the sample spinner
  • The plot shows the sample data, the data constructed from the chosen PCs and the residual (sample - reconstruction)


  • No labels

1 Comment

  1. ---When the calculation is complete, the % variance explained should be plotted against the component number in the Variance Explained tab.

    Can you explain what would be expected in the plot if there are 2 or 3 components present? (i.e. What does the y-axis mean in practise)

    ---The Scores/Loads tab allow the difference PC scores to be plotted against each other

    typo

    ---The plot on the left shows the average spectrum/pattern plotted with the loadings vector for the selected PCs

    The loadings vector means the variation from the mean? If there is a spike in the line then that part of the selected PC is changing (at some point) in the series?