(將資料轉(zhuǎn)換成數(shù)字形式).ppt

1、1,Analysis of Quantitative Data,Social Research Methods 2113 & 6501 Spring, 2007 March 26April 2, 2007,2,Quantitative Analysis: convert data to a numerical form and statistical analyses,quantification (量化): the process of converting data to a numerical format (將資料轉(zhuǎn)換成數(shù)字形式),3,Part I. Dealing With Data

2、,(過錄)編碼簿製作、原始資料編碼 (a codebook & coding data) 資料輸入 (entering data) 清理資料 (cleaning data),4,Coding Data (資料編碼),Data Coding: systematically reorganizing raw data into a format that is machine readable (將原始資料有系統(tǒng)地轉(zhuǎn)化成電腦程式可閱讀的形式) Need a coding procedure & a codebook Coding procedure(編碼程序): a set of rules st

3、ating that certain numbers are assigned to variable attributes (一些設定的規(guī)則，以確定變數(shù)屬性的編號為何),5,Coding Data (資料編碼),A codebook(編碼簿): a document describing the coding procedure and the location of data for variables in a format that computers can use (一份描述編碼程序及變數(shù)位置的文件) Should be well-organized, detailed (有條理、

4、詳細) Make multiple copies! (多影印幾份，妥善保存) Precoding (事前編碼) done Code each questionnaire (每份問卷都需編碼),6,A codebook example,7,Codebook Example,8,請問影響您繼續(xù)升學的主要原因：(可複選) 1_a1 目前所就讀的科系有很大的發(fā)展前途 1_a2 為了滿足父母的期待 1_a3 家庭經(jīng)濟足以支持您繼續(xù)升學 1_a4 個人成績優(yōu)異 1_a5 對從事學術研究有興趣 1 = yes, 0 = no 1_a6 目前就業(yè)率不高 8 = skip, 9 = missing 1_a7 目

5、前就業(yè)市場不需所就讀科系之人才 1_a8 為了延後兵役 1_a9 繼續(xù)升學可獲得較佳的升遷機會 1_a10 繼續(xù)升學可獲得較高的薪水 1_a11 個人認為取得碩士文憑以上是重要的 1_a12 社會普遍升學風氣興盛 1_a13 因為同學都升學 1_a14 老師的建議 勾選完後請將您認為最重要的三個原因圈選出來 1mi_1, 1mi_2, 1mi_3 88 = skip, 99 = missing,9,編碼注意事項:,製作編碼簿 (a codebook) 記得加上兩個變數(shù): 編碼者號碼及受訪者ID Ex: 變數(shù)名稱coder & ID 製作SPSS資料檔 輸入變數(shù)名、 變數(shù)註解、 數(shù)值及註解、遺漏

6、值等 請用紅筆編碼，綠筆用來訂正 編碼者請在問卷左上方或右上方簽名 資料輸入者也請在問卷封面上簽名,10,Entering data (資料輸入),Data records: each row (列) represents a respondent (case) each column (行): a specific variable Four ways to enter data: Code sheet Direct-entry method Optical scan Bar code Recommend: double entry (重複輸入),11,Cleaning Data (清理資料

7、),Why clean your data? Need accuracy (準確度)! Double data entry to ensure data accuracy. Procedures to “clean” data (清理資料的步驟): Use frequency tables to do code cleaning (利用次數(shù)分配表檢查變數(shù)屬性數(shù)值、遺漏值等) Contingency cleaning (consistency checking) (一致性檢查) EX: cross-classifying two variables Check contingency quest

8、ions (檢查條件式問項答案) Will discussed in more details later,12,Part II. Data Analysis,13,Review: Quantitative Data Analyses,Univariate analyses (單變量分析): a single variable Distributions, central tendency, variation Bivariate analyses (雙變量分析): the analysis of two variables The scattergram, percentaging a ta

9、ble, contingency tables Multivariate analyses (多變量分析): analyzing more than two variables simultaneously,14,Please note: some calculations are not suitable to all variables,Continuous variables (ratio variables) (連續(xù)變數(shù)): a variable whose attributes form a steady progression Ex: age Discrete variables

10、(間斷變數(shù)): a variable whose attributes are separate from one another, or discontinuous Ex: gender Nominal or ordinal variables Can use raw numbers, percentages, or modes (but not very interesting),15,Need a research plan for quantitative data analysis:,Think about your research question(s) & research p

11、urposes (先想想你的研究問題及目的) 記得: 分析資料不能偏離主題! Think about your hypotheses (想想你的假設) Main variables: dependent variables & independent variables (依變數(shù)及自變數(shù)是什麼?) Propose a plan for data analysis, timeline, hours of work, etc. (擬定計劃，如何進行及分工),16,A doable process for data analysis,Data Cleaning (will discussed in

12、detail later) Understand your data Look at frequency tables of key variables again (再看一次重要變數(shù)的次數(shù)分配表) Find any potential problems for further data analysis (ex: missing data percentage?) (尋找問題: 遺漏值太多?其他? Produce some tables for descriptive analysis (製作表格，注意表格格式) Have some sense for data recording (想想變

13、數(shù)如何重新編碼),17,Data Analysis: data recoding and create new variables,Why data recoding and crate new variables? For more meaningful analyses Ex: dummy variables, collapse response categories Data recoding and editing: Think about your hypotheses, have purposes Remember to check consistency after record

14、ing (use frequency tables) (新舊變數(shù)次數(shù)分配要一致) Add variable and value labels (新變數(shù)要有變數(shù)及數(shù)值註解),18,Bivariate analyses (雙變量分析),確定你的研究性質(zhì) 建議: 想想你的依變數(shù)，如果是名義或次序量數(shù)，只能用cross-tabulation與卡方來分析相關性。 依變數(shù)如果是等距或等比量數(shù)，可採用複迴歸。自變數(shù)要先分析變數(shù)之間的共線性。,19,Bivariate Tables,Contingency tables: values of the dependent variable are conting

15、ent on (depend on) values of the independent variables How to percentage a table? Note: the direction of percentaging in tables is arbitrary In general, we follow this “tip”: “row” by “column” (dep. var by indep. var) Usually done by SPSS of other computer programs How to read a table of percentages

16、?,20,21,Reading a percentage table:,A rule of thumb (the tip): if the table is percentaged down, read across (表格是直行百分比，依橫列來解讀) If the table is percentaged across, read down,22,Bivariate Tables without Percentages,Use the mean in a bivariate table (i.e., subgroup comparison),23,An example of continge

17、ncy table analysis:,Q: 性別與大學主修之關係,24,卡方值(2)是17.825，自由度是1，由表得知此卡方值大於10.827 (p = 0.001)， 所以得到此卡方值的機率應低於0.001， 觀察到的性別與大學主修之間的關係不可能只是因抽樣誤差而產(chǎn)生。 由列聯(lián)表中可看出，約半數(shù)的男性選擇念人文社會或自然科學， 但女性僅有略多於四分之一的比率(26.6%)在大學時選擇主修自然科學， 性別與大學主修之差異在統(tǒng)計上是顯著的。,25,An example of data analysis: 教育成就的世代差異,Q: 什麼因素可以解釋孩子的教育程度? 資料:第四期第三次的臺灣社會

18、變遷基本調(diào)查(於2002年完成) 注意: 此範例大幅簡化了因果模型,26,Data Analysis Example,27,Data Analysis Example,重新編碼 受訪者的教育程度、受訪者父親的教育程度、受訪者母親的教育程度這三個變數(shù): (無) 0 (自修) 3 (小學) 6 (國、初中、初職) 9 (高中普通及職業(yè)科、高職、士官學校) 12 (五專、二專、三專、軍警校專修班) 14 (軍警官學校、技術學院、科技大學、大學) 16 (碩士) 18 (博士) 20,28,Data Analysis Example,29,30,31,Data Analysis Example,由於受訪者父親與受訪者母親的教育程度有相當高的相關性，因此兩者可以互相解釋，將此兩變項同時引入複迴歸的方程式中，將導致違反迴歸方程式的基本假設:兩自變項 Xi 、Xj 無共線性。 在選用上以遺漏值較少的受訪者母親教育程度為其變數(shù)。在迴歸方程式處理遺漏值時，無論是取消此一數(shù)值，或者以平均值代替遺漏值，或其他的處理方式，都會以人為方式影響到迴歸方程式的計算，因此，在避免人為甘預的原則下，我們會採用遺漏值越少的變數(shù)。,