Another way of representing the outcome of the San Francisco Airport 2011 survey is by Principal Component Analysis. The plot below is a bit cluttered, but you can see the outcome 'overall' at about four o'clock.
The more aligned the other variables are with 'overall', then the greater the agreement (or correlation) between them. So 'screens' and overall quality are highly correlated, whereas car rental is not. So the CEO should do two things:
*make sure that screens and escalators are kept to a high standard
*look into shuttle, car rental etc and see if the lack of strong correlation means something. The variables that are pointing to the north-east are in a group that has 'travel to the airport' in common. Perhaps it is tough getting to the airport, but once you're there all is good.
SFO1 <- read.csv("C:/Users/Stephen/Dropbox/Consulting/smartphone/SFO1.csv",
row.names = 1)
require(FactoMineR)
sfair <- PCA(SFO1, graph = FALSE)
plot(sfair, choix = "var")