Beyond the basics (mean, variance, distributions, simple test, regression, ...) statistics looks like a huge mess -to me- . Any method has typically 10 variants and can be 'colored' in 10 flavors, not to mention technicalities like missing data handling, binary data, ... and people's names I need to think a long time before I understand the main point of a method (say the bartlett test in a PCA). I never feel confident for "the details and the options". Many things never get translated in simple words. Mathematical descriptions are not (easily) available or are (practically) unreadable. Sometimes I have the feeling that a "huge monte carlo" point of view could help to simplify the understanding. For example, a statistical test could be implemented by a "monte carlo" method. If computers had a huge computing power, recourse to specific tests (hard to develop, get and understand) would be useless. At least, this could be a language to express a "statistical problem". Could some of you give me some related ideas, papers or web sites? I am equally interrested by nicely-written material, organized references, unification attempts. I would like to see the forest, not the trees.