本研究將許春金、馬傳鎮等於1997~1999年對板橋、新莊地區建檔以及2007年追蹤刑案記錄整合資料之401位犯罪青少年調查資料,以早期偏差,毒品有無,偏差價值觀,偏差友伴,遊樂型休閒,是否父母同住等六個變項做為自變項,以「再犯有無」做為依變項,進行邏輯斯迴歸。發現除了第4(偏差友伴)與第5(遊樂型休閒)的假設未成立外,早期偏差,毒品有無,偏差價值觀,以及是否父母同住等四個變項,皆能預測是否再犯。其中以毒品有無的影響最為重要。接下來將有再犯者依其犯罪生涯、再犯次數與再犯時距進行分類,可分得三群:偶發犯罪者、短暫犯罪者與持續犯罪者。而這三群在犯罪生涯、再犯次數與再犯時距三方面的表現,均有明顯的不同。進一步以早期偏差,毒品有無,偏差價值觀,偏差友伴,遊樂型休閒,是否父母同住等六個變項做為自變項,以集群分析所得的偶發犯罪者、短暫犯罪者與持續犯罪者三群做為依變項進行判別分析。則可發現所導出的判別函數式,對於此樣本的判別有63.9%的正確率,也顯示判別函數的準確度。在兩個判別函數中,以第一個函數具有顯著的解釋力或預測力,而其中相對重要性較高的有毒品有無與早期偏差。本研究發現:早期偏差,毒品有無,偏差價值觀,以及是否父母同住等四個變項,皆能預測是否再犯。而其中以毒品有無的影響最為重要。並對於犯罪青少年的處遇,以及預防再犯提出以下建議:應加強虞犯青少年的輔導;應強化反毒與戒治工作;應強化家庭功能。
This study use the data of Sheu & Ma's research, the data is survey to 401 Juvenile delinquency samples who lived in Banqiao and Xinzhuang City, Taipei County during 1997~1999. The Logistic regression shows the independent variables as early delinquency behaviors, drug using, deviation values, and live with parents or not, can forecasting the dependent variable: repeat an offense or not. Then we use the cluster analysis to classify with the crime career, times of reoffence, and the duration of reoffence, we can get three types of offenders: accidental offender, brief offender, and continue offender. Then we use the discriminant analysis, we same use early delinquency behaviors, drug using, deviation values, deviation peers, recreational activities and live with parents or not to be the independent variables, the crime career, times of reoffence, and the duration of reoffence to be the dependent variable. The discriminant shows the 63.9% correct rate.The study finds that the early delinquency behaviors, drug using, deviation values, and live with parents or not, can forecasting the delinquency repeat an offense or not, especially drug using. About correcting with delinquency youth, and preventing reoffence, we suggest that should enhance the guidance of delinquency teen-agers; enhance the anti drug and rehabilitation works; and enhance the family function.