本研究主要目的在檢視「特殊教育長期追蹤資料庫」失敗樣本對推論可能造成偏頗的影響因素。以該資料庫民國96學年度學前及國小三歲、五歲、小一和小三等四組4,873名有效樣本資料爲對照,追蹤調查隨機抽取之527名失敗樣本在性別、居住地區、父親教育程度、母親教育程度、障礙類別、安置型態、障礙影響程度、家長參與程度及學習進步程度等九個變項上的情形,以卡方考驗和多變量邏輯斯迴歸分析進行失敗樣本和有效樣本在這些標竿指標上的差異考驗。分析結果發現,四組失敗樣本與有效樣本之間的差異主要在家長參與程度、障礙影響程度和學習進步程度三項,其中,尤其以家長參與程度爲最大關鍵因素。
The purpose of this study was to examine nonresponse bias in a large-scale survey, the Special Needs Education Longitudinal Study, which interviewed special needs children's parents, teachers and school administrators. Data analyses included a comparison between the originally selected sample and the follow-up survey for unit-nonresponses on the demographic characteristics and variables related to learning outcomes. The results revealed that except for the 5-year-old sample, there was no significant difference between these two groups on demographic variables for the 3-year-old, the first-grade, and the third-grade samples; on the contrary, the variables related to learning outcomes were the main causes accounting for the nonresponses. Multivariate logistic regression models were applied to estimate the odds of correlates of nonresponse, including demographic characteristics and variables related to learning outcomes. Again, the results demonstrated that parental involvement, a variable related to learning outcomes, was a significant predictor of nonresponse bias across 3-year-olds to 3rd graders. None of the demographic characteristics was statistically significant. For the pre-school stage, except for parental involvement, influence of disabilities was another index to predict nonresponse. For the elementary stage, learning progress was an additional key variable to portend nonresponse. The findings suggest that the variables related to learning outcomes, rather than demographic characteristics, may be the important sources of nonresponse bias in the Special Needs Education Longitudinal Study.