• Communications for Statistical Applications and Methods
• /
• v.6 no.3
• /
• pp.755-760
• /
• 1999

In multivariate stratified sampling surveys it is general to use a few stratification variables which are highly correlated with the important variables at design stage. But there might be some secondary study variables which are not so highly correlated with those stratification variables. In that case it is not efficient to use the same type of estimator due to the secondary variables as the one base on the important variables. A post-stratified estimation is proposed to increase the efficiency of the estimator with existence of secondary variables. The proposed method is illustrated with a set of fishery household population survey data.

• The Korean Journal of Applied Statistics
• /
• v.36 no.6
• /
• pp.547-559
• /
• 2023

Proper handling of nonresponse in sample survey improves the accuracy of the parameter estimation. Various studies have been conducted to properly handle MAR (missing at random) nonresponse or MCAR (missing completely at random) nonresponse. When nonresponse occurs, the PSA (propensity score adjusted) estimator is commonly used as a mean estimator. The PSA estimator is known to be unbiased when known sample weights and properly estimated response probabilities are used. However, for MNAR (missing not at random) nonresponse, which is affected by the value of the study variable, since it is very difficult to obtain accurate response probabilities, bias may occur in the PSA estimator. Chung and Shin (2017, 2022) proposed a post-stratification method to improve the accuracy of mean estimation when MNAR nonresponse occurs under a non-informative sample design. In this study, we propose a double post-stratification method to improve the accuracy of the naive PSA estimator for MNAR nonresponse under an informative sample design. In addition, we perform simulation studies to confirm the superiority of the proposed method.

• Survey Research
• /
• v.12 no.3
• /
• pp.173-186
• /
• 2011

We introduced the methodologies used to construct the longitudinal weights and cross-sectional weight that are required for the analysis of Korea Youth Panel Survey. To analyze the longitudinal dynamic change of the population, we derived the longitudinal weight through nonresponse adjustment based on logistic regression and post-stratification. Cross-sectional weights that are necessary to produce an asymptotically unbiased estimator of the population parameter were constructed through simple nonresponse adjustment based on overall response rate and post-stratification.

• Journal of Korean Society of Forest Science
• /
• v.99 no.3
• /
• pp.285-291
• /
• 2010

Since 2006, the $5^{th}$ National Forest Inventory (NFI) has been implemented to provide forest resources statistics at the national level and at the county level as well. However, it needs a small-area estimator for estimating forest statistics at the county-level due to a small number of samples collected within a county. This study was conducted to evaluate the applicability of a geographical-based synthetic estimator for estimating forest growing stock volumes at the county level. The NFI-field plots surveyed were post-stratified into three forest cover types. In the synthetic estimator, field plots within a geographical-based super-county for each county were used to estimate stratum weights and stratum mean volumes. It was resulted that estimated stratum weights using the synthetic estimation were significantly differ from forest cover maps. The standard errors of estimated mean by the synthetic estimation that ranged from ${\pm}3.5\;m^3$/ha to ${\pm}7.7\;m^3$/ha were more smaller than those (${\pm}7.8\;m^3/ha{\sim}{\pm}24.7\;m^3/ha$) by the direct estimation. This means that the synthetic estimation is possible to provide more precise estimates of mean volumes.

• Survey Research
• /
• v.7 no.2
• /
• pp.53-69
• /
• 2006

Weights can be made and imposed in both sample design stage and analysis stage in a sample survey. While in design stage weights are related with sample data acquisition quantities such as sample selection probability and response rate, in analysis stage weights are connected with external quantities, for instance population quantities and some auxiliary information. The final weight is the product of all weights in both stage. In the present paper, we focus on the weight in analysis stage and investigate the effect of such weights imposed on the weighted mean when estimating the population mean. We consider a finite population with a pair of fixed survey value and weight in each unit, and suppose equal selection probability designs. Under the condition we derive the formulas of the bias as well as mean square error of the weighted mean and show that the weighted mean is biased and the direction and amount of the bias can be explained by the correlation between survey variate and weight: if the correlation coefficient is positive, then the weighted mein over-estimates the population mean, on the other hand, if negative, then under-estimates. Also the magnitude of bias is getting larger when the correlation coefficient is getting greater. In addition to theoretical derivation about the weighted mean, we conduct a simulation study to show quantities of the bias and mean square errors numerically. In the simulation, nine weights having correlation coefficient with survey variate from -0.2 to 0.6 are generated and four sample sizes from 100 to 400 are considered and then biases and mean square errors are calculated in each case. As a result, in the case or 400 sample size and 0.55 correlation coefficient, the amount or squared bias of the weighted mean occupies up to 82% among mean square error, which says the weighted mean might be biased very seriously in some cases.