چکیده:
Stratified sampling design is used in many different sampling surveys. The method which has being used since many years for stratifying frame members is Dalenius-Hodges stratification method which strongly depends on the distribution of data. This paper contains a study of the former method and Fuzzy clustering method as a method that does not impose any constraint on the distribution of data. Moreover، a comparison between these two methods is considered. As a result، the conclusions of the study shows that in skewed data Fuzzy clustering method is more efficient than the Dalenius-Hodges Stratification method for skewed data and it can be used in surveys at Statistical Center of Iran (SCI).
خلاصه ماشینی:
57-69 Comparison of the Dalenius-Hodges Stratification Method with Fuzzy Clustering in Skewed Data Arman Bidarbakhtnia, {o*o}Ma'roufi Unit Statistical Center of Iran Abstract.
Section 3 is dedicated to reviewing fuzzy clustering, and in Section 4, results related to the comparison of stratifications obtained by applying the two methods on real data with different distributions and two sets of data generated from normal and uniform distributions are presented.
As is evident from the method, fuzzy clustering, unlike the Dalenius-Hodges method, does not make any assumptions regarding the distribution of data; rather, the basis of this method is the similarity values between observations in a cluster, which are obtained based on a distance function.
To compare the Dalenius-Hodges classification methods with fuzzy clustering, two groups of data were used.
4-Results of Comparing the Methods The results of the comparisons show that, as expected, since the Dalenius-Hodges classification method always seeks to keep the h S h N values equal among strata as much as possible, in skewed data, it significantly affects the efficiency of the sampling design.
(refer to page image) As can be seen in Table 1, the average intra-stratum standard deviation value for the classification resulting from the fuzzy method is less than that of the Dalenius-Hodges method for all data.
Examination of the heterogeneity criterion shows that the fuzzy clustering method significantly reduces the heterogeneity criterion within strata compared to the Dalenius-Hodges classification method (Table 2).
(
, 
, 
)
