Part of an answer in http://www.statistik-forum.de/versuchsplanung-f36/bestimmung-der-trennscharfe-t8731.html

Example of CART in rpart with only nominal predictors

Screeplot of a principal components analysis of a 12 item questionnaire based on roughly 2000 people.Uploaded for a question on www.statistik-forum.de

Normal Q-Q-Plots are not easy to judge. What is "straight enough"? Someone on www.statistik-forum.de wanted to test a n=19 sample for normality. This gives a number of n=19 truely normally distrbuted "Normal-Q-Q-Plots" for reference.