maxlike classification problem

Discussion created by campgl01 on Jun 16, 2014
Hello, i dont know if someone could help me with this...I am doing a cluster classification using maxlike and isocluster for create the signature field. After the classification i calculate the confusion index by doing the difference between the assignation  probability of a pixel to the belonging class and the assignation probability to the second more probable class. This probabilities are calculates using class probability tool.  I have estandarized all the variables and i have done two classification at different escales...But i am having problems with classification at a third resolution scale (180m), it is not correct even following the same criterio as i did in the previous classification, since the classification is not logic and when i calculate de confusion index, there are pixels for which the class probaility is lower than that of the second class more probable and that doesn't make sence.
Has someone had the same problem using maxlike, obteining wrong classifications using the same criteria than in a previous right classifcation??

Many thanx