我國各級學校師資之預測

Abstract

本研究以時間數列分析來預測民國80-85學年度各學校教師數，包括幼稚園，國民小學，中學（包括國中及高中，但不包括高職），專科學校，及大學（含大學部、碩士班及博士班）。也順便預測就讀大專院校學生佔年底人口的千分比，年中人口數，及出生數，以供經建計劃單位參考。由於高職教師數經一次差分後，acf，inverse acf，及pacf皆沒有一期達顯著，無法以單變項的時間數列分析來預測，故以轉換函數分析來嘗試，以高職學生數數列為自變項。方法仍根據Box & Jenkins(1976)為主。以政大電算中心之SAS/ETS的套裝程式執行。預期預測誤差三年內不超過5%。此種預測可供教育計劃之參考。 This study used time-series analysis to forecast the teach number in different levels of education in Taiwan: The number of teachers in kindergarten, elementary school, middle school, junior college, and university. Additionally, the number of students studying in the higher educational institutions per thousand end-year population, the number of mid-year population, and the number of live-birth were also forecasted. Because of the fact, that no significant lag in the acf, inverse acf, as well as pacf were found after first order differencing by the series of teacher number of vocational senior high school, no suitable order of autoregressive and moving average process could be suggested for forecasting. Therefore, the transfer function analysis was used instead of the univariate time series analysis. The analysis method for both the univariate time series analysis and the transfer function analysis is chiefly based on the Box and Jenkins (1976) employed by the SAS/ETS statistical package. Hopefully, the forecast error would not exeed the 5% threshold ((real value － predicted value)/real value). The forecast of teacher number will find its meaning, if the results of the forecast are taken into account in the educational planning.