S = 44.31 R-Sq = 76.9% R-Sq(adj) = 72.2% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 65190 32595 16.60 0.001 (回归方程显著)
Residual Error 10 19636 1964
Total 12 84826
Source DF Seq SS
G1/K 1 56191
G1/G2 1 8999
Unusual Observations(异常观察值)
Obs G1/K 8"EWT Fit SE Fit Residual St Resid
12 0.122 1300.0 1380.9 25.5 -80.9 -2.24R
R denotes an observation with a large standardized residual
 40”EWT与G1/K、G1/G2的回归分析
Regression Analysis: 40"EWT versus G1/K, G1/G2
The regression equation is
40"EWT = 6363 - 22880 G1/K - 11883 G1/G2
Predictor Coef SE Coef T P
Constant 6362.7 893.9 7.12 0.000
G1/K -22880 6605 -3.46 0.006
G1/G2 -11883 4752 -2.50 0.031
S = 50.35 R-Sq = 74.3% R-Sq(adj) = 69.1% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 73139 36570 14.42 0.001 (回归方程显著)
Residual Error 10 25356 2536
Total 12 98495
Source DF Seq SS
G1/K 1 57287
G1/G2 1 15852
Unusual Observations(异常观察值)
Obs G1/K 40"EWT Fit SE Fit Residual St Resid
2 0.113 1632.0 1697.8 40.0 -65.8 -2.16R
R denotes an observation with a large standardized residual
 EKCO与G1/K、G1/G2的回归分析
Regression Analysis: EKCO versus G1/K, G1/G2
The regression equation is
EKCO = 208 - 1215 G1/K + 83 G1/G2
Predictor Coef SE Coef T P
Constant 207.63 30.56 6.79 0.000
G1/K -1215.4 225.8 -5.38 0.000
G1/G2 82.7 162.5 0.51 0.622
S = 1.721 R-Sq = 76.0% R-Sq(adj) = 71.2% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 93.774 46.887 15.82 0.001 (回归方程显著)
Residual Error 10 29.629 2.963
Total 12 123.403
Source DF Seq SS
G1/K 1 93.006
G1/G2 1 0.768
Unusual Observations(异常观察值)
Obs G1/K EKCO Fit SE Fit Residual St Resid
2 0.113 82.500 84.758 1.369 -2.258 -2.16R
R denotes an observation with a large standardized residual
(3) 结论
回归分析结果显示,回归方程拟合良好,但三个回归方程中均出现了异常观察值,这对回归方程的拟合性有一定的影响。在后续的工作中需要对回归方程的拟合性进行优化。
【改进阶段】
(1) 优化各参数回归方程的拟合程度
由于观察值“2”的试验结果在40”EWT和EKCO的回归方程中均出现了异常报警,说明该组试验数据的可靠性存在一些问题,我们决定将该组异常数据去除后重新拟合回归方程,结果如下:
 8”EWT与G1/K、G1/G2的回归分析
The regression equation is
8"EWT = 6410 - 37231 G1/K - 2952 G1/G2
Predictor Coef SE Coef T P
Constant 6409.9 700.7 9.15 0.000
G1/K -37231 7582 -4.91 0.001
G1/G2 -2952 4408 -0.67 0.520
S = 37.30 R-Sq = 84.8% R-Sq(adj) = 81.5% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 70097 35048 25.19 0.000 (回归方程良好)
Residual Error 9 12520 1391
Total 11 82617
Source DF Seq SS
G1/K 1 69473
G1/G2 1 624
 40”EWT与G1/K、G1/G2的回归分析
The regression equation is
40"EWT = 7030 - 39748 G1/K - 4156 G1/G2
Predictor Coef SE Coef T P
Constant 7030.0 729.1 9.64 0.000
G1/K -39748 7890 -5.04 0.001
G1/G2 -4156 4587 -0.91 0.389
S = 38.81 R-Sq = 86.2% R-Sq(adj) = 83.1% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 84602 42301 28.08 0.000 (回归方程显著)
Residual Error 9 13557 1506
Total 11 98159
Source DF Seq SS
G1/K 1 83365
G1/G2 1 1237
 EKCO与G1/K、G1/G2的回归分析
The regression equation is
EKCO = 231 - 1794 G1/K + 348 G1/G2
Predictor Coef SE Coef T P
Constant 230.51 24.86 9.27 0.000
G1/K -1793.8 269.0 -6.67 0.000
G1/G2 347.6 156.4 2.22 0.053
S = 1.323 R-Sq = 85.4% R-Sq(adj) = 82.1% (回归方程拟合良好)
Analysis of Variance
Source DF SS MS F P
Regression 2 91.813 45.907 26.22 0.000 (回归方程显著)
Residual Error 9 15.756 1.751
Total 11 107.569
Source DF Seq SS
G1/K 1 83.160
G1/G2 1 8.653
Unusual Observations
Obs G1/K EKCO Fit SE Fit Residual St Resid
12 0.117 76.900 79.731 0.490 -2.831 -2.30R
R denotes an observation with a large standardized residual
结论:异常观察值剔除后,回归方程拟合程度更加良好。
(2) 利用优化后的回归方程进行预测,并确定最佳G1/K、G1/G2间隙值
根据整管电特性的要求,EKCO最佳值为80V左右,8”EWT最佳值为1400~1500μA,40”EWT最佳值为1600μA左右。根据这些要求,我们利用回归方程对G1/K、G1/G2进行预测,结果如下: