Randomized algorithm for adjusting sample data

A. V. Orekhov

Vestnik Sankt-Peterburgskogo universiteta. Prikladnaâ matematika, informatika, processy upravleniâ, no. 3 (2015), pp. 96-104

Résumé

One of the main problems of statistical analysis of experimental data is obtaining an unbiased (representative) sample. It is a natural desire to obtain a representative sample of computational methods. The procedure for adjusting the structure of the sample in accordance with the structure of the population is said to be “adjusted sample”. All known methods of adjusting the sample data have a significant drawback: these methods are “correct” empirical distribution functions, not the sample. In this paper we introduce the concepts of control spaces and studied traits, given the formal definition of a representative sample population; described randomized, algorithm, adjustment is sample data, not their empirical distributions. Iteration of algorithm of randomized sample corrections of data can be described as follows. Please select one spectral value of the control characteristics such that the absolute difference between the respective selective and general shares has a maximum value, then have all $k\mid 1\le k\le n$,  sought $\max|w_k-p_k|$. If spectral values of a few, then randomly selects any of them. Let it be the $i$-th component of the vector $\mathbf {W}^{X}$ and $\mathbf {P}^{X}$. Then, if $w_i-p_i>0$, of the total sample randomly is to remove some vector $\mathbf{Z_t}$, in which the component $x_i=1$, if $w_i-p_i0 $, the total sample randomly duplicated vector $\mathbf{Z_t}$, in which the component $x_i=1$. Refs 9.

Détail

Keywords: randomized algorithm, sampling frame.