what n and p denotes in n*p input matrix of pca?

I m bit confused matrix which is to be fed as input to princomp should be in the form of 160*25 or 25*160 where 25 are coffecients and 160 are features. as when I give input as 25*160 it is generating coefficient matrix and with 160*25 generating error Eigen value<0. Kindly help me.

thanx

Answer 1

[Walter Roberson]

"Rows of X correspond to observations, columns to variables"

Features would correspond to variables, but it is not obvious what you mean by "coefficients".

I would expect there to be problems if you implicitly told it there were more variables than there were observations, by using an array that was taller than it was wide, so I would expect that you need to pass in the 25 x 160 matrix.

[parul mittal]

sr 25 coefficients means input images that I m using and 160 are the features that I have extracted per image. if I give 25*160 matrix as input to pca than its output coefficient matrix will be 160*160. than how can I use it as testing matrix for classifier when my target matrix will be 5*25, error will b generated target and testing mismatch. kindly help me thnx