# Some clarifications on eigenvectors and eigenvalues from PCA

Could somebody tell me whether suggestions in bold true or not?

Chapter 2.2 Interpretation of the eigenvectors/eigenportfolios


This paper says that loadings in the maximal eigenvector need to be all positive and should not change sign, what if i have negative ones, can i force them to be positive always by simply taking them by module, e.g. MathAbs(Vector) ?

Q # 2 : The same paper also defines weights for eigenportfolio in this way :

Q[i] = EigVecCoef[i] / StdDev[i]


There is also another paper that says that eigenvector is an angle (or direction) of the portfolio's spread which allows to map current portfolio's spread to initial axes (dimensions) :

http://georgemdallas.wordpress.com/2013/10/30/principal-component-analysis-4-dummies-eigenvectors-eigenvalues-and-dimension-reduction/

So i do not understand - why do i need to divide each value in eigenvector by standard deviation to calculate weights if this portfolio is already mapped to initial axes?

-