To probe into the possibility on survey instead of census

• Main Author: 楊宗龍
• Other Authors: 余清祥
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/81684220787074986468

碩士 === 國立政治大學 === 統計研究所 === 102 === Census aims to collect the information of the population, including its size, structure, and other related information, in a country or region. Census data are used widely, including policy planning for the government, business strategies for private sectors, or individuals who are interested in their living environment. In order to reduce the rising non-response rate and survey cost, some countries are looking for the possibility of introducing new census methods, such as registered-based census and rolling census. Taiwan also conducted the first registered-base census in 2010 and about 16% of people were asked to fill in the census form. This is very different from the previous full population census and we are interested in knowing if 16% of population is sufficient to provide accurate estimate of the whole Taiwan population. The objective of this study is thus to explore how many people shall we survey (in percentage of population) to acquire reliable estimate of the total population. In particular, we are interested in the sex and age structures of population, in addition to the population size. To simplify the discussion, we shall only consider the simple random sampling since more complicated sampling plan (e.g., two-stage or clustered sampling). Also, we assume that the sampling is without replacement and the chi-square test is applied to check if the sample estimate of population structure is the same as the true population. There are two types of evaluation criteria: relative and absolute errors. We found that 16% of sample can only provide reliable population estimate up to age 80 (5-age group: ages 0-4, 5-9, …, 75-79, 80+).