الفهرس يوجد فقط 14 صفحة متاحة للعرض العام
 |  | from | 120 |  |  |
| | | from | 120 | | |
المستخلص
Household surveys are one of the most important tools that are used by national statistical agencies, social scientists and market researchers to measure characteristics of human populations to find required estimates. In order to Maximize and extracted the results from the sample to population is used so-called sampling weights. Frequently proper analysis requires the use of sampling weights to reduce or eliminate the estimation bias. But sometimes sampling weights variable has a high dispersion for several reasons, for example: Disparity population count from one region to another. By high non-response by the respondents in some areas. Or not full coverage of all sampling units. Large variation in survey weights can cause inferential problems for data users. Trimming large weights methods have been developed to adjust such extreme weights. Thus, there is a need for diagnostic measures of the impact of weight trimming on survey inference that extend past the existing 2design effect3 type of summary measures. In this dissertation, the proposed research goals are to use trimming weights strategy and to evaluate the effect of the weight trimming in terms of the sampling variance and design effect