Basic interpretation of PCA plots

[#1]

Hi all, I was wondering if someone can explain the second graph that shows up after general PCA analysis on JMP. There are three graphs that typically appear: eigenvalues chart, a second scatterplot, and the loading graph. What's the relationship between the second graph and the loading graph? I notice the loading graph typically has a lot more details than that second scatter plot. Thanks!

[#2]

There are many ways that PCA can be explained, but here is one explanation.

The second graph is the scatterplot of PC1 and PC2. This is a plot of the observations on the new PC coordinate axes. You could think of it as a two-dimensional view of your original multidimensional space. It can tell you about relationships between the observations.

The third graph is the plot of the eigenvalues for the first two PCs. This can tell you about relationships between the variables. 

The two together: suppose you see an outlying point in the upper right portion of the second graph. Look at the third graph and see which variables point in that direction. That outlying point is likely high in those variables. Similarly, any variables in the opposite direction would have low scores for the original variables. 

[#3]

Thank you Dan that was very helpful!

[#4]

@crystal007 there is many excellent Information in the program on this, e.g. have a look at

• some examples in the statistic index (Help --> statistics index), or in other sample tables
• in the report press "?" and Point to the part of report you have Questions about
• in the Manual (multivariate Methods, chapter 4 PCA)
• ...

[#5]

Thank you Georg that was also very helpful! It helped me derive the biplot view from the PCA plot and loading graph.