试验计划法DOE.PPTVIP

Three DOE Class 90a 實驗結果 統計圖表分析 假設檢定_Fixed Effects Model 若拒絕H0 ，則不同之因子水準對反應變數有影響。 反之，則無影響。 ANOVA表格 其中：a 為因子之水準數 n 為每一水準之資料個數(複製次數) 決策模式 SST = SSE + SSTreatments ， 且SST有an-1個d.f. ， 則SSE /s2及 SSTreatmentss2皆為Chi-square隨機變數，其d.f.分別為a(n-1)及a-1 。(Cochran’s Theorem) 若 F0 Fa,a-1,a(n-1) ，則不同之因子水準對反應變數有影響。 反之，則無影響。 a 為相對風險。 Example 1：布料強度之研究 模式參數之估計 Unbalanced Data If the number of observations for each treatment is different (unbalanced data), then Example 2：紙張強度之研究 殘值分析 (Residual Analysis) 確定殘差(Residual)來自於自然變異N(0,s2) Residual Analysis I – Normality Plot Standardized Residuals Residual Analysis II Residual Vs. Factor Levels (Treatment) = Equal Variances 資料轉換(Data Transformation) 若Residual Vs. Estimates之圖有任何異常時，應考慮做資料轉換。 常用之資料轉換公式： Poisson Data = Square Root Transformation Lognomal Data = Logarithmic Transformation Binomial Data = Arcsin Transformation Empirical Selection of a Transformation (pp. 83-86) Tests for Equality of Variance Bartllet’s Test (pp. 81) Modified Levene Test (pp.82) Used when normality assumption is possibly violated. Comparisons Among Treatment Means(pp.86-104) Graphical Comparisons of Means Constrasts Orthogonal Constrasts Scheffe’s Method (Comparing all Constrasts) Comparing Pairs of Treatment Means Turkey’s Test Fisher’s LSD Method Duncan’s Multiple Range Test Newman-Keuls Test Dunnett’s Procedure (Comparing to a constant) Sample Computer output (pp. 105) C.V. = (s.d. / mean) * 100% PRESS: Predict Error Sum of Square R-Squared = SSTreatments / SST Adj R-Squared =1-[SSE/(N-a)]/[SST/(N-1)] Pred R-Squared = (SST-PRESS)/ SST Adeq Precision must greater than 4. 分析程序 樣本大小 (pp. 107~110) Operating Characteristic Curves Specifying a Standard Deviation Increase Confidence Interval Estimation Method Miscellaneous Topics (pp. 110~119) Discovering Dispersion Effects The Regression Approach to ANOVA Nonparametric Methods in ANOVA Kruskal-Wallis Test Rank Transformation The Variance Model 當因子之全部可能水準（水準個數較多時）皆為研究之範圍時，吾人可利用Variance Mode