Intra-class correlation and ANOVA for multi-class classification?

Hello

I have trained and evaluated with leave-one-out cross-validation classifiers (SVM, logistic regression, deep belief network, gaussian naive bayes, k-nearest neighbour) on my dataset. My dataset consists of 4 classes where each class has around 10 examples (10 feature vectors).

I have found the following Matlab library to calculate the intra-class correlation coefficient: http://ch.mathworks.com/matlabcentral/fileexchange/22099-intraclass-correlation-coefficient--icc-?s_tid=srchtitle

There are 6 types of ICC. How can I use ICC in my case and what does it tell it to me?

Second, how can I use ANOVA for calculating statistic in my case? Can I calculate the p-value with ANOVA?

Answer 1

[Sepp]

Does somebody have a hint? Would really appreciate it.