Скачать или смотреть Factoshiny and R Commander: Case study 9: Red wine PCA

Since wine quality cannot be attributed using just one physicochemical variable, it could be done using all the variables at the same time (PCA). Students built the loading and the score plot using Factoshiny. They observed correlations between variables in the score plot such as the inverse correlations between alcohol content and density and the inverse correlation between pH and citric acid.
The loading plot shows variations with high impotence in the PCA model according to their color and correlations between variables. For example, as expected, fixed acidity and citric acid were inversely correlated to pH, while free sulfur and total sulfur dioxide were highly related.
In the score plot students may observe that high-quality red wines have more ethanol content and less total sulfur, and free sulfur dioxide than low-quality wines.
The primary learning objective was to demonstrate how PCA can be used to visualize a large data set. Students must recognize the most discriminating variables in the loading plot (Figure 3) and the importance of original variables in sample distribution through the score plot (Figure 4).
As a learning model, students can use PCA to visualize the data sets. For example, they could not find any information by just looking at spreadsheet 7. However, they visualized correlations between variables in the loading plot and the relationships of each physicochemical variable with wine quality in the score pot. Additionally, students claimed that Factoshiny provided top quality images, where samples were represented in distinct colors according to their qualitative variables.