统计与信息论坛

规模以下工业调查所提供的经营活动指标对于工业统计至关重要。国家现行方案依据最新经济普查数据构建双重抽样框，从中抽取固定样本连续调查多年，以构造非经济普查年规模以下工业经营活动指标的HT估计量。目前，规模以下工业单位流动性高且变动频繁，将直接影响固定样本对非经济普查年目标总体的代表性。因此，提出在非经济普查年，采用基于双重抽样框的分层二重抽样方法。每年度先抽取一个大样本用于估计规模以下工业总体的新增率、消亡率以及规模大小。再利用一个较小的第二重样本进行实际调查，并结合由第一重样本所得数据优化的抽样权数，最终给出规模以下工业总体总量估计。推导其抽样偏差和方差，构造基于刀切法的方差估计量。针对非经济普查年规模以下工业营业收入总额的估计，验证了新方法相较于现行固定样本HT估计量的精度更高。该研究有助于提高非经济普查年规模以下工业调查经营活动指标估计的可靠性，并为相关调查方案设计提供理论依据。

The business activity indicators provided by industrial surveys under a designated size are of crucial for industrial statistics.The current national scheme adopts a dual sampling frame based on the latest economic census data to select a fixed sample for continuous survey multiple years, in order to construct a HT estimator for indicators of industrial activities under a designated size in non-census years.At present, the high liquidity and frequent fluctuations of industrial units under a designated size will directly affect the representativeness of the fixed sample to the non-census years target population.This article proposes a stratified double sampling method based on double sampling frame in non-census years.This method requires that the sample be taken again every year.The overall approach begins by drawing a large sample to estimate the growth rate, extinction rate, and size of the industrial units below a designated size.Subsequently, using the actual survey data from a smaller second-phase sample, combined with sampling weights optimized based on information obtained from the first-phase sample, estimates for the population of the industrial units below the designated size are provided.Derive its sampling bias and variance, and construct a variance estimator based on the knife cut method.The new method has been validated to have higher accuracy compared to the current fixed sample HT estimation for estimating the total industrial operating income under a designated size of non-census years.This research contributes to improve the reliability of economic index estimation and provides theoretical basis for related survey scheme design.