SAS / PC
十四、 範 例 十四 : 兩組平均值間差異檢定 ( PROC TTEST ) 及變方分析 ( PROC ANOVA )
程式檔名稱 : SAMPLE14
|
|
DATA TTEST; | 平均值間 t 值檢定, @@ |
INPUT TREAT $ Y @@; | 為橫向輸入處理及觀測值 |
CARDS; | (處理僅為兩組) |
A 1.94 A 3.26 A 2.70 A 3.21 A 3.30 A 2.07 A 2.10 | |
B 1.73 B 1.94 B 1.91 B 1.69 B 2.08 B 1.77 B 2.48 | |
; | |
RUN;
|
|
PROC TTEST; | |
CLASS TREAT; 宣告處理別欄位名稱 | |
VAR Y; 進行兩組別間, Y 欄位資料之 t 值檢定 | |
RUN; (表 14-1) | |
駢對式 t 值檢定 | |
DATA PIREDT; 輸入新資料集 | |
INPUT BEFORE AFTER; | |
DIFF=AFTER-BEFORE; BEFORE 試驗開始之心跳次數 | |
CARDS; AFTER 試驗結束之心跳次數 | |
72 102 DIFF 個體試驗前後之心跳次數差異 | |
80 85 | |
81 111 | |
78 102 | |
83 140 | |
72 82 | |
69 90 | |
; | |
RUN;
|
|
PROC MEANS MEAN STDERR T PRT MAXDEC=3; 檢定試驗前、後之差異值是否為零, | |
VAR DIFF ; | 並將差異平均值印出 (表 14-2.1) |
PROC MEANS MEAN STDERR MAX MIN N MAXDEC=3; | 計算試驗前、後心跳次數之平均值 |
VAR BEFORE AFTER; | |
RUN;
|
(表 14-2.2)
|
PROC ANOVA ; 變方分析 | |
兩組或兩組以上之平均值比較 | |
PROC ANOVA DATA=TTEST; | 變方分析 (利用原有 t 值檢定資料) |
CLASS TREAT; | 宣告處理欄位 |
MODEL Y=TREAT; | 分析模式的設定 |
應變數(DEPENDS)=自變數(INDEPENDS) | |
MEANS TREAT | 那些處理欄位作進一步平均值間差異比較 |
/ DUNCAN; | 選擇以 DUNCAN 方法來作比較 (ALPHA= |
0.05; DEFAULTED) (表 14-3) | |
MEANS TREAT / LSD TUKEY | 選擇以 LSD 及 TUKEY 方法來作比較, 設 |
ALPHA=0.01 CLDIFF; | 定 ALPHA=0.01 (即採納了對立擬說時,但 |
RUN; | 其為錯誤, 此種機率需小於 0.01 你才認 |
定為差異顯著) 並估計信賴區間 (CLDIFF) | |
不同試驗設計之模式設定及 F 值檢定範例 | |
|
如 下頁所示
|
表 14-1 | |
TTEST PROCEDURE | |
Variable: Y | |
TREAT N Mean Std Dev Std Error | |
------------------------------------------------------------------------------------ | |
A 7 2.65428571 0.61285825 0.23163865 | |
B 7 1.94285714 0.27311257 0.10322685 | |
Variances T DF Prob>|T| | |
--------------------------------------------------------- | |
Unequal 2.8053 8.3 0.0223 | |
Equal 2.8053 12.0 0.0159 | |
For H0: Variances are equal, F' = 5.04 DF = (6,6) Prob>F' = 0.0699 | |
不同試驗設計之模式設定及 F 值檢定範例 |
|
********************************************************* | |
* EXAMPLE OF RANDOMIZED COMPLETE-BLOCK DESIGN * 完全逢機區集設計 | |
* MODEL : Yijk = u + Bi + Tj + eijk * | |
* PROC ANOVA; * | |
* CLASS BLOCK TREAT; * | |
* MODEL Y=BLOCK TREAT; * | |
* MEANS BLOCK TREAT / DUNCAN; * | |
* RUN; * | |
* * | |
********************************************************* |
|
* EXAMPLE OF FACTORIAL EXPERIMENT * 複因子試驗 | |
* MODEL : Yijk = u + F1i + F2j + (F1*F2)ij + eijk * | |
*----------------------------------------------------------------------------------------- * | |
* PROC ANOVA; * | |
* CLASS FACT1 FACT2; * | |
* MODEL Y=FACT1|FACT2; * | |
* MEANS FACT1 FACT2/ DUNCAN; * | |
* RUN; * | |
********************************************************* | |
* EXAMPLE OF LATIN SQUARES DESIGN * 拉丁方格設計 | |
* MODEL : Yijkl = u + Ri + Cj + Tk + eijkl * | |
*----------------------------------------------------------------------------------------- * | |
* PROC ANOVA; * | |
* CLASS ROW COLUMN TREAT; * | |
* MODEL Y=ROW COLUMN TREAT; * | |
* MEANS ROW COLUMN TREAT /DUNCAN; * | |
* RUN; * | |
********************************************************* | |
* EXAMPLE OF NESTED DESIGN * 巢式設計 | |
* MODEL : Yijk = u + Si + Dj(i) + eijk * | |
*----------------------------------------------------------------------------------------- * | |
* PROC ANOVA; * | |
* CLASS SIRE DAM; * | |
* MODEL Y=SIRE DAM(SIRE); * | |
* MEANS SIRE; * | |
* TEST H=SIRE E=DAM(SIRE); * | |
* RUN; * | |
********************************************************* | |
* EXAMPLE OF SPLIT-PLOT DESIGN * 裂區設計 | |
* MODEL : Yijkl = u + Bi + F1j + (B*F1)ij + F2k * | |
* + (F1*F2)jk + eijkl * | |
*------------------------------------------------------- * | |
* PROC ANOVA; * | |
* CLASS BLOCK FACT1 FACT2; * | |
* MODELY=BLOCK FACT1BLOCK*FACT1 FACT2 FACT1*FACT2; * | |
* TEST H=BLOCK E=BLOCK*FACT1; * | |
* TEST H=FACT1 E=BLOCK*FACT1; * | |
* MEANS FACT1 / DUNCAN E=BLOCK*FACT1; * | |
* RUN; * | |
*************************************************************; | |
|
|
表 14-2.1 | |
Analysis Variable : DIFF | |
N Obs Mean Std Error T Prob>|T| | |
7 25.286 6.391 3.957 0.0075 | |
表 14-2.2 |
|
N Obs Variable N Minimum Maximum Mean Std Error | |
-------------------------------------------------------------------------------------------------------- | |
7 BEFORE 7 69.000 83.000 76.429 2.034 | |
AFTER 7 82.000 140.000 101.714 7.492 | |
--------------------------------------------------------------------------------------------------------- | |
表 14-3 |
|
Analysis of Variance Procedure | |
Dependent Variable: Y | |
Sum of Mean | |
Source DF Squares Square F Value Pr > F | |
Model 1 1.77145714 1.77145714 7.87 0.0159 | |
Error 12 2.70111429 0.22509286 | |
Corrected Total 13 4.47257143 | |
R-Square C.V. Root MSE Y Mean | |
0.396071 20.64063 0.47444 2.2985714 | |
Source DF Anova SS Mean Square F Value Pr > F | |
TREAT 1 1.77145714 1.77145714 7.87 0.0159 | |
Duncan's Multiple Range Test for variable: Y | |
NOTE: This test controls the type I comparisonwise error rate, not the | |
experimentwise error rate | |
Alpha= 0.05 df= 12 MSE= 0.225093 | |
Number of Means 2 | |
Critical Range 0.551 | |
Means with the same letter are not significantly different. | |
Duncan Grouping Mean N TREAT | |
A 2.654 7 A | |
B 1.943 7 B | |
. | |
. 略 |