Hi All

Assume that I have to classify one image by a Neural Network 5 times (5 iterations) and each time it gives me these results:


Iteration 1 : class1: 0.0, class2: 0.5, class3: 0.5, class4: 0.0 >> The total of the classes in this iteration is ONE.
Iteration 2 : class1: 1.0, class2: 0.0, class3: 0.0, class4: 0.0 >> The total of the classes in this iteration is ONE.
Iteration 3 : class1: 0.5, class2: 0.5, class3: 0.0, class4: 0.0 >> The total of the classes in this iteration is ONE.
Iteration 4 : class1: 0.0, class2: 1.0, class3: 0.0, class4: 0.0 >> The total of the classes in this iteration is ONE.
Iteration 5 : class1: 0.0, class2: 0.0, class3: 1.0, class4: 0.0 >> The total of the classes in this iteration is ONE.




I need to do a statistical test to know which CLASS is significantly different from the other classes.


My four sets will be:
1- class1 values from all iterations = (0.0, 1.0, 0.5, 0.0, 0.0)
2- class2 values from all iterations = (0.5, 0.0, 0.5, 1.0, 0.0)
3- class3 values from all iterations = (0.5, 0.0, 0.0, 0.0, 1.0)
4- class4 values from all iterations = (0.0, 0.0, 0.0, 0.0, 0.0)




I thought to use "One-way ANOVA" then "Tukey's test" exactly like here: Using One-way ANOVA and Tukey?s test to compare data sets | Clever Owl


But the problem is that my data is kind of dependent.You can see that each iteration is added up to 1.
For example, in iteration#1, 0.0 + 0.5 + 0.5 = 1. In iteration#2, 1.0 + 0.0 + 0.0 = 1 and so on.


So, which is the best statistical test I can use in my situation?


Thank you