大数据背景下的抽样问题探讨

发布时间：2019-06-22 09:13

【摘要】：作为从十九世纪末就逐渐得到认可并延续至今的抽样调查技术,不仅仅在获取数据和统计推断方面发挥了作用,而且代表了一种在保证数据结构不变的情况下将数据以大化小思想。但是,在大数据时代下,不断改进的存储设备和计算能力让人们拥有了取之不尽、用之不竭的数据资源,还有更加接近结果的预测方法,人们似乎无需再设定总体,更不用再根据抽样调查的设计方案在总体中抽取样本,取而代之的是具有全数据形式的大数据。所以有越来越多的人开始忽略抽样调查,认为大数据时代下抽样调查已经没有存在的意义。大数据时代是不是就不需要抽样了?大数据背景下抽样面临哪些挑战?如何做好大数据时代的抽样?本文的写作是围绕这样一系列问题而进行的探索和思考,文中首先从抽样调查在诞生的历史背景出发,指出抽样调查是在实践中为了用少量数据代表总体数据而产生的,并明确了抽样的学科地位其发展过程中为调查工作所发挥的价值。然后,结合大数据时代的特点和抽样调查自身的不足,说明了大数据时代下抽样调查面临的挑战,包括抽样调查中的非抽样误差和抽样误差的存在,还有对子总体估计的困难性等。接着,为了阐述大数据时代下抽样调查在仍然有其存在的必要性,对大数据、总体、大样本和全数据等概念进行了辨析,从概念和定义上表明了大数据既不是抽样调查中所说的总体,也不能代表全数据;同时对大数据和抽样数据的来源和特征以及在数据分析方法和思路上的差别进行了深入剖析。最后,基于完善大数据时代的抽样,使抽样与大数据有效地衔接,配合发展的目的,本文探讨了两个方面的问题,一是抽样调查数据与大数据之间的融合和补充,二是把随机抽样调查技术和非随机抽样调查技术分别应用到分析大数据的方法和过程中。
[Abstract]:As a sample survey technology that has been recognized and extended to the present since the end of the nineteenth century, it is not only a role in the acquisition of data and statistical inference, but also represents a small idea that will be used to make the data structure unchanged. However, under the age of large data, the continuous improvement of the storage and computing power has made people have the inexhaustible data resources, and the prediction method of closer results, people don't seem to need to set the whole, Instead of taking samples from the overall sample, the sample is typically extracted from the sample, instead of large data in the form of full data. So there are more and more people to ignore the sample survey, and it is believed that the sample survey in the big data age has not existed. Isn't there a need for sampling in the big data age? What are the challenges to sampling in the context of large data? How to make the sampling of the big data age? The writing of this paper is to explore and think about a series of problems, first, from the historical background of the birth of the sample survey, it is pointed out that the sample survey is produced in practice in order to represent the overall data with a small amount of data, The value of the research work in the course of its development is clarified. Then, combined with the characteristics of the large data age and the shortcomings of the sample survey, the challenges faced by the sample survey in the large data age are described, including the non-sampling error in the sample survey and the existence of the sampling error, and the difficulty in the overall estimation of the sub-population. Then, in order to explain the necessity of the sample survey in the age of large data, the concept of large data, general, large sample and full data is analyzed. At the same time, the source and characteristics of the large data and the sampling data and the difference between the data analysis method and the thought are analyzed. At last, based on the perfection of the sampling of the big data age, the sampling and the big data are effectively connected, and the purpose of the development is discussed in this paper. The first is the fusion and the supplement between the sample survey data and the big data. The second is to apply the random sampling technique and the non-random sampling technique to the method and the process of analyzing the big data, respectively.
【学位授予单位】：兰州财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C81

【参考文献】