<<Previous, 1, 2, 3, 4, 5, 6, 7, 8, 9, Next >>
Principal Component Analysis (PCA)
Principal Component Analysis (PCA) attempts to identify the underlying factors that explain the pattern of correlations within a set of observed variables and samples. Understanding of similarity or dissimilarity of samples using extracted factors is much easier than using unprocessed data.
When you click "Open Form" at the sheet “Multibase_Preparation", Navigator will open again.
Click "PCA" to open PCA dialog box.
In this dialog box, you can select variable or sample names shown on PCA maps. In this tutorial, click "Done" without any changes.
Three new sheets, Multibase_PCA, Multibase_PCA2 and Multibase_Result will be generated. Score (sample) and loading (variable) scatter plots are displayed in Multibase_PCA, Goodness of fit (R2), prediction ability (Q2) and distance to the model (DmodX) are shown in the sheet of Multibase_PCA2, and the calculated values of each component are shown in the Multibase_Result. If more than three principal components are extracted, three sets of scatter plots, [PC1 and PC2], [PC1 and PC3], and [PC2 and PC3] are shown in the sheet of Multibase_PCA.
The score plot shows the similarity or dissimilarity of samples (countries), and the loading plot represents summary of the variables (foods and drinks). The two plots are complementary and can be superimposed. The direction of the loading plot corresponds to the same direction in the score plot. The score plot (right graph) shows how the 27 country’s consumption profiles relate to each other. Countries close to each other have similar properties, whereas countries far from each other have dissimilar consumption profiles. The result that Asian countries (China, Japan and Korea) being located together in the left-hand side of fist component (PC1) represent similarity in food and drink consumption. Hungary, Poland and Norway are located in the center of the map, which indicates that they have average properties.
The loading plot shows which variables are influential for the model and how the variables are correlated to each other. The dots of whiskey and beer are close, which means the beer consumption profile is similar to whiskey one. When the value of whiskey consumption increases or decreases, the value of beer consumption will change in the same manner.
Asian countries located in the left-hand side of PC1 means that their whiskey consumption is less than European or American countries.
If you would like to know dot's name, you simply move the mouse cursor over the target dot. The variable or sample name is displayed in the score and loading plots.
If you need more information, you can get them in Multibase_Result sheet.
"Contribution" means the value which each component can explain for original data. Forty nine percent contribution (component 1) means 51 % of original information is lost and component 1 represents about half of original data. The contribution of second component is 21% and accumulated contribution of comp1 and comp2 goes up 69%. This means that the scatter plot between PC1 and PC2 covers 69% of original data. The more the components used, the more the accumulated contribution gets close to 100%. However, note that the increasing of components decrease reliability of model. Multibase extracts components whose relative eigenvalues are larger than one.
In previous study, you knew that the profile of food and drink consumption in Asia is quite similar. However, the score plot you saw was quite incomprehensible because you could not understand which dot represents Japan, or Asia if you did not move the cursor on it.
Multibase can display categorization by colored ellipses to make your interpretation easy. In order to categorize samples, you go back DataSheet1 sheet, and click “Multibase2015” and “Open Form”.
When the dialog box opens, variable and sample name area you set before will be inherited, because Multibase memorizes previous actions. If you want to clear histories, click “Reset” button, then Multibase erases all histories.
To categorize countries, click “Samples” tab.
2-3-1 Marunouchi Chiyoda Tokyo Japan