Sample Selection Method Research Based On Multivariate Spatial Sampling Design

The research stagnated where the sampling method is based on the field of the single target variable,and since then the theory of spatial sampling has been put forward in the mid and late twentieth Century.Therefore,we studied the sampling method both of the spatial multivariate non probability sampling and of the multivariable space stratified sampling.On the basis of the form sampling method,we set the space position as a unique marker of target variables and then establish the functional relation among variables and overall goal to ensure the minimum variance and unbiased sample mean.For the multivariable space stratified sampling method,according to the correlation among the target variables and the latitude or longitude,we design the functional relation to stratify.There are six parts in this article.We firstly introduce the research background of multivariate spatial sampling theory.Then,this paper states the basic assumption and the related statistical knowledge on multivariate spatial non-probabilistic sampling method.This method's hypothesis is that auxiliary variables are independent for each other;besides,two more restrictions are sample mean unbiased and minimum variance of mean estimator.On the condition of these,we establish the relation among the target variables and spatial geographical position and find out the selection method of optimal samples.Furthermore,we studied the sampling method of multiple-objective space stratified sample.We suppose correlations among the overall target,longitudes,latitudes and target variables exist significantly.Based on the above preconditions,a quadratic transformation is applied to transform target variables into a single target population;in addition,we set the relation between this single target population and spatial geographical location information.Clearly,this relation is on longitudes,latitudes and the target variables;and using the character,which is that one function's derivative is the turning point,to stratify the overall population.This article,which guaranteed both unbiased estimation of sample mean and minimum variance in sampling process,puts forward a new method of multivariate spatial non-probabilistic sampling method.The creativity come from the following.This method,firstly,can efficiently control multivariate deduction error so as to improve sampling efficiency.Furthermore this method,compared with traditional sampling methods,help us to get the minimum variance.On the other hand,this paper explain to us how to stratify layers of multivariate spatial stratified sampling method.This method,contrasting with traditional ones,regards spatial geographical position as a variable in laying function and guarantee sampling efficiency by minimize discrepancies on the same layer and maximize discrepancies between different layers.Innovativeness of the stratified sampling method is that both multivariate similarity and spatial proximity of units on the same layer can be considered in this method.