...

When visualizing high throughput data, for example via a 2D or 3D scatter plot, to display the distance of the observations on NGS observations where each observation has thousands of measurements, dimension reduction techniques need to be used. One can apply Principle Component Analysis, t-Distributed Stochastic Neighbor Embedding etc. Those techniques convert data from high dimensional space to lower dimensional space conveying similar information. When information represented in lower dimensional space is not exactly the same as the one from higher dimensional space data, error is generated. In the advanced option settings of dialogs like PCA, t-SNE, there is Generate mapping error statistics option, check checking this button will output the mapping error on the plot.

...