統(tǒng)計學-計量資料

1、2024/3/21,Xi'an Jiaotong University-Medical School,第二章 計量資料的統(tǒng)計描述,集中趨勢描述 均數(shù):對稱分布,特別是正態(tài)分布資料 幾何均數(shù):等比級數(shù)資料,對數(shù)正態(tài)分布資料 中位數(shù):不適合用幾何均數(shù)描述的偏態(tài)分布資料 末端有不確定數(shù)據(jù)的正態(tài)分布資料離散趨勢描述

2、 極差:不穩(wěn)定 四分位數(shù)間距:用于偏態(tài)分布資料 方差與標準差:用于正態(tài)分布資料,描述平均變異\絕對變異 變異系數(shù):用于正態(tài)分布資料變異度對比\相對變異,2024/3/21,Xi'an Jiaotong University-Medical School,例2-1 從某單位1999年的職工體檢資料中獲得101名正常成年女子的血清總膽固醇的測量結果如下，試

3、編制頻數(shù)分布表。,2024/3/21,Xi'an Jiaotong University-Medical School,編制步驟,求極差 極差（range）也稱全距 確定組段數(shù)和組距 根據(jù)組距寫出組段 分組劃記并統(tǒng)計頻數(shù),,,2024/3/21,Xi'an Jiaotong University-Medical School,表2-2 115名正常成年女子血清轉(zhuǎn)氨酶（mmol/L）含量分布,2024/3/21

4、,Xi'an Jiaotong University-Medical School,2024/3/21,Xi'an Jiaotong University-Medical School,頻數(shù)表和頻數(shù)分布圖的用途,描述頻數(shù)分布的類型 描述頻數(shù)分布的特征 便于發(fā)現(xiàn)一些特大或特小的可疑值 便于進一步做統(tǒng)計分析和處理,2024/3/21,Xi'an Jiaotong University-Medical Schoo

5、l,,,2024/3/21,Xi'an Jiaotong University-Medical School,集中趨勢的描述 –均數(shù),直接計算法頻數(shù)表法,,,2024/3/21,Xi'an Jiaotong University-Medical School,集中趨勢的描述 –幾何均數(shù),公式,,,,2024/3/21,Xi'an Jiaotong University-Medical School,集中趨勢

6、的描述 –中位數(shù),n為奇數(shù):n為偶數(shù):,,,,,,,,2024/3/21,Xi'an Jiaotong University-Medical School,離散趨勢描述-極差\四分位距,極差R: 最大值與最小值之差四分位距 = -,,,,,,2024/3/21,Xi'an Jiaotong University-Medical School,離散趨勢描述-方

7、差\標準差,總體方差與總體標準差樣本標準差,,,,,,2024/3/21,Xi'an Jiaotong University-Medical School,離散趨勢描述-變異系數(shù),不同均數(shù)資料的變異度比較不同度量單位資料的變異度比較,,2024/3/21,Xi'an Jiaotong University-Medical School,一、常用相對數(shù),強度相對數(shù)：率 某時期內(nèi)發(fā)生某現(xiàn)象的觀察單位

8、數(shù) 同期可能發(fā)生某現(xiàn)象的觀察單位總數(shù)結構相對數(shù) ：構成比 某一組成部分的觀察單位數(shù) 同一事物各組成部分的觀察單位總數(shù)相對比 甲指標/乙指標,,,2024/3/21,Xi'an Jiaotong University-Medical School,二、計算相對數(shù)的注意事項,結構相對數(shù)不能代替強度相對數(shù)（以比代率）計算相對數(shù)應有足夠數(shù)量正確計

9、算合并率注意資料的可比性相對數(shù)的抽樣誤差,2024/3/21,Xi'an Jiaotong University-Medical School,SPSS軟件簡介,常用窗口:數(shù)據(jù)窗\程序窗\結果輸出窗口變量定義: 變量名:不超過8個字符,最好以英文字母加數(shù)字組成,避免用中文 變量類型:數(shù)值型\日期型\字符型 小數(shù)位數(shù) 缺失值定義 變量標簽:可用中文 值標簽:分類變量需要定義值標

10、簽文件類型: 數(shù)據(jù)文件.sav 程序文件.sps 結果文件.spo,2024/3/21,Xi'an Jiaotong University-Medical School,產(chǎn)生新變量,數(shù)據(jù)排序:sort,2024/3/21,Xi'an Jiaotong University-Medical School,數(shù)據(jù)庫建立,設計正確的數(shù)據(jù)庫，保證統(tǒng)計軟件能識別數(shù)據(jù)庫一般為行列表第一行為變量名一

11、個case 占用一行一個variable占一列盡可能采用數(shù)據(jù)形式錄入,2024/3/21,Xi'an Jiaotong University-Medical School,數(shù)據(jù)的前處理,數(shù)據(jù)庫:數(shù)據(jù)的儲存\備份\使用權限 數(shù)據(jù)雙輸?shù)谋匾?數(shù)據(jù)庫名稱,變量定義,必要的說明 原始數(shù)據(jù)的重要性!!!電子數(shù)據(jù)庫\紙質(zhì)數(shù)據(jù)庫數(shù)據(jù)考察:數(shù)據(jù)類型? 分布?

12、異常值? 缺失數(shù)據(jù)的處理?數(shù)據(jù)清理與鎖定數(shù)據(jù)庫軟件: Access, EpiData, Oracal 統(tǒng)計軟件中的數(shù)據(jù)庫,2024/3/21,Xi'an Jiaotong University-Medical School,描述性統(tǒng)計,計量資料 圖表形式：箱式圖、直方圖 指標形式：集中趨勢：均數(shù)、中位數(shù)、幾何均數(shù)

13、 離散趨勢：極差、四分位距、方差、 標準差、變異系數(shù)分類資料 圖標形式：直方圖、線圖 指標形式：率、比,2024/3/21,Xi'an Jiaotong University-Medical School,數(shù)據(jù)庫一般操作,數(shù)據(jù)庫合并：merge 追加變量、追加病例數(shù)據(jù)庫拆分：split衍生數(shù)據(jù)：

14、產(chǎn)生新的變量,2024/3/21,Xi'an Jiaotong University-Medical School,常用命令_衍生數(shù)據(jù)命令,產(chǎn)生新變量 (transform) Recode: numeric --- string Compute: constant functions Date/time:,2024/3/21,Xi'an Jiaotong U

15、niversity-Medical School,常用命令_數(shù)據(jù)步命令1,數(shù)據(jù)步 Insert variables Insert case Go to case Sort cases,Select cases: give some conditionsWeight cases: just used for grouped database.,2024/3/21,Xi'an Jiaotong Unive

16、rsity-Medical School,常用命令_數(shù)據(jù)步命令2,File commands Split file: split file into sub-database according to the categorical variables Merge files: Add cases: the files should be have same structures. Ad

17、d variables: Key variable (unique identification number for the observations),2024/3/21,Xi'an Jiaotong University-Medical School,常用命令_數(shù)據(jù)步命令2,數(shù)據(jù)庫架構改變Transpose: row --- column (transpose rows and

18、 columns ) Restructure: long --- short (repeated measurement),2024/3/21,Xi'an Jiaotong University-Medical School,描述性統(tǒng)計過程,Descriptive statistics Frequencies: continuous variables

19、 categorical variables Descriptive: continuous variables (Normal distributed) Explore: continuous variables Crosstabs: categorical variables,2024/3/21,Xi'an Jiaotong Un

20、iversity-Medical School,基本統(tǒng)計圖,Graphs Bar graph: continuous/categorical variables (comparison) Pie graph: categorical variables (constituent ratio) Line graph: pr

21、oportions and time variables (time trend) Histogram: continuous variables (frequency distribution) Scatter plot: continuous variables (cor

22、relation analysis ) Box plot: continuous variables (check median, quartiles, maximum, minimum),2024/3/21,Xi'an Jiaotong University-Medical School,統(tǒng)計圖_連續(xù)型變量,measures that are recorded on a natur

23、ally occurring number scale.Box plotsHistogram,2024/3/21,Xi'an Jiaotong University-Medical School,Graph commands,HistogramHorizontal scale represents the value of the variableThe vertical scale represents the fre