- 學位論文
主成分分析於SDQ問卷之應用併以SDQ探討外籍配偶子女心理健康狀態
The Application of Principal Component Analysis in the Strengths and Difficulties Questionnaire (SDQ) and exploring the Mental Health Status of Foreign Spouses` Children via SDQ
摘要
目的：中文版SDQ(Strengths and Difficulties Questionnaire)之適用性與其信效度分析。使用主成分分析制訂簡化中文版SDQ問卷。以SDQ探討台灣外籍配偶家庭子女與非外籍配偶家庭子女的心理健康狀態是否有差異，並試圖找出其他可能影響心理健康狀態之危險因子。 方法：選取彰化縣的彰化市、二林鎮、大城與芳苑鄉為城市、小鎮、鄉村三類行政區，以概率比例抽樣方式，隨機抽出1250位小學學童進行家長版、老師版、學生版SDQ問卷評量。描述三版本問卷原始分數分布狀態，以百分位切點方式，決定問卷的正常值、邊緣值、異常值分數切點，計算外籍配偶家庭子女精神疾患高危險群盛行率。信效度方面，以Cronbach`s alpha評估問卷內部一致性，以皮爾森積差相關評估三版本間的施測者關聯性。試圖用主成分分析分析問卷結構，其25個問題在試驗族群代表變異量為何，並從適當個數的主成分中選出足以代表原始資料大部分變異的子群變項做為簡化版問卷。最後，將SDQ的1.整體困難分數為連續資料2.簡化版問卷分數為連續資料 3.整體困難分數是否落於異常值為二元資料，三類型的結果個別當依變項。將相關因子，包括性別、年級、行政區、父母是否為外籍配偶、父母教育程度、父母有無離婚、有無與父母同住、有無兄弟姊妹、家庭年收入視為自變項，以混合線性模型進行分析。 結果：共收集到有效問卷，老師版本和家長版本各1008份，學童版本356份。建立常模與問卷正常值、邊緣值、異常值分數切點結果及計算外籍配偶家庭子女精神疾患高危險群盛行率。三種版本各別的內部一致性高，除了在同儕關係面向上，家長和學童版本的alpha僅36%與33%。版本間的兩兩相關性不高，反應三種觀測者對樣本觀察結果的認知方式不相同，且給分的高低皆有顯著差異。主成分分析則分別選出〈1〉家長版本的6與13個問題〈2〉老師版本的5與9個問題〈3〉學童版本的6與13個問題做為迷你版與簡化版SDQ。最後，以混合模型分析，得到比複回歸模型更好的配適度。除了在考量了長處項目的簡化版問卷得到東南亞與其他國籍子女有較差的心理健康狀態外。不論結果設為連續或是二元變項下，在使用原始問卷統計下，外籍配偶子女皆和本國籍配偶子女心理健康狀態無統計學上的顯著差異。然則，鄉村、男生、低家庭年收入、是否與父母同住，則是造成1.整體困難分數升高2.較差簡化版問卷心理健康狀態3.容易掉入〈兒童青少年精神疾患高危險群〉的危險因子。 結論： SDQ的簡化版與迷你版本之建立，證實具有以少數簡化問題即能代表原始變項在台灣國小學童施測的好處，可做為大規模篩檢的工具。在使用原始版本SDQ的整體困難分數當結果，不管是二元或是連續變項結果，外籍配偶家庭子女與本國籍配偶家庭子女心理健康狀態並無差異。但需注意的是，在納入長處項目的簡化版問卷，東南亞與其他國家外籍配偶子女有較差的心理健康狀態。
並列摘要
Objectives：We established the norms and evaluated the reliability and validity of SDQ in elementary school students in Taiwan. We then used principal component analysis to reduce the original variables of SDQ in order to yield a simplified version of SDQ. Finally we compared the mental health status using SDQ between foreign spouses’ children and non-foreign spouses’ children and also tried to identify the possible risk factors that influence the mental health status. Methods：We selected three regions of Changhua county, including Changhua City(city), Erhlin Town(town), and Dacheng and Fangyuan Country(country) in the light of probability proportionate to size(PPS) sampling. We collected 1250 elementary school students to complete the parent, teacher, and self-report versions of SDQ by their parents, teachers and themselves. The distributions of original scores of SDQ were specified. We determined the cutoff of normal, borderline and abnormal scores of SDQ and calculated the prevalence of high risk group of children and adolescent mental disorder in foreign spouses’ children. We assessed the internal consistency of SDQ via Chronbach’s alpha and the inter-rater correlation via pearson’s product-moment correlation coefficient. We used principal component analysis to analyze the structure of SDQ and to assess the variations of the 25 questions of SDQ in our study group. We selected adequate numbers of subset original variables that capture the items accounting for large variation to make a simplified version of SDQ. Finally, we applied the linear mixed model to assess the mental health status between foreign spouses’ children and non-foreign spouses’ children and other possible risk factors, including gender, grade, region, mother’s and father’s education level, the martial status of their parents , living with their parents or not, having siblings or not, family annual income. We regarded three kinds of outcome as dependent variables, the scores of total difficulties of SDQ ,the scores of simplified SDQ and the probabilities of entering high risk group of children and adolescent mental disorder, and those factors mentioned above as independent variables. Results：We collected valid SDQ questionnaires, including teacher version for 1008 copies, parent version for 1008 copies and student version for 356 copies. We described the norms , the cutting points of normal, borderline and abnormal scores of SDQ and the prevalence of high risk group of children and adolescent mental disorder in foreign spouses’ children. The three versions of SDQ revealed high internal consistency except the domain of peer problems in parent and student versions. The inter-rater correlations were so low as to reflect different viewpoints of the three observers and the different scores between the three versions. We selected 6 and 13 questions of parent version, 5 and 7 questions of teacher version and 6 and 13 questions to represent mini- and simplified versions of SDQ via PCA. Finally, we analyzed the data via the linear mixed model and reached better goodness of fit. Except the finding that the group of ‘east-south Asia and others’ got poorer scores of mental health status via simplified version of SDQ, we found there were lack of obvious differences between foreign spouses’ children and non-foreign spouses’ children in their mental health status in another two assessments. The region of country, male, low family annual income and living with their parents or not were the risk factors responsible for higher scores of total difficulty of SDQ, poorer scores of mental health status via simplified version of SDQ and high risk group of children and adolescent mental disorder. Conclusions：The mini- and simplified versions of SDQ can capture lots of variations of mental status of elementary school students in Taiwan via less questions and is helpful for comprehensive screening. By taking the scores of total difficulty of SDQ as the outcome, the mental health status between foreign spouses’ children and non-foreign spouses’ children was lacking of difference . However, we paid attention to the finding that the group of ‘east-south Asia and others’ yielded poorer scores of mental health status via simplified version of SDQ which considered the aspects of strengths.