In PLS/PCR we can get a RMSE-components curve curve to decide the number of components. Similarly, can we obtain such a curve for N-finder analysis to decide the number of endmembers for the data?
(https://i.stack.imgur.com/o8eHe.png)
I could get the endmembers for my data. But there is always ambiguity in deciding the end members. What could be the best way to decide the number of endmembers is not known to me.