Shape analysis Flashcards
What is Principle component analysis
PCA is a technique designed to summarize large amounts of data into something simpler, while maintaining the important info
able to reduce the number variables in a dataset and extract new factors that explain variation in the data.
used to generate relative warps
What are relative warps
they’re relative shape changes in landmark configurations along principal warps in GM analyses.
they outline localised shape changes and demonstrate where shape changes are most intense summarizing overall shape changes among tested configurations.
eigenvalue
eigenvalues measure the amount of variation in the total samples accounted for by each principal component.
The ratio of eigenvalues is the ratio of importance of the factors with respect variation in the raw variables.
ie If a factor has a low eigenvalue, then it explains little variances in the raw data (can be ignored as considered redundant to higher eigen value PCs)
PC scores:
PC scores are computed as vectors of deviations of the observations from the sample mean, multiplied by the vectors of PC coefficients,
on a graph the axis are two PC values