표집의 설계에 따른 표준오차 비교 분석 연구

초록

The purposes of this study were to illustrate the mismatch between replicated sampling and traditional analytic model based on simple random sampling, in terms of standard error and to compare traditional analytic model and replicated sampling by using empirical data. The major findings were as follows; First, ignoring the sample design led to biased and misleading estimates of the standard error. In other words, ignoring the sample design resulted in substantially underestimating the standard error. Second, it appeared that the standard errors of unstandardized regression coefficients were substantially different by way of dividing units in each stratum into two random groups. Finally, Special methods and softwares were recommended to avoid biases that arise when statistical software assumes the data are independent and identically distributed.