Sample size and statistical power considerations in high-dimensionality data settings: a comparative study of classification alg