Analyze Variability

Coefficients Table - Effects

  

Look at the effects in the coefficients table to determine the relative strength of the effects.

Use these guidelines to analyze the effects:

·    The absolute value of the effect determines its relative strength. The higher the value, the greater the effect on the response.

·    The signs of the effects indicate which factor settings reduce the variability of the response.

-    A positive sign indicates that the low factor setting results in lower variability than the high setting.

-    A negative sign indicates that the high factor setting results in lower variability than the low setting.

Example Output

 

Coded Coefficients for Ln(Std)

 

                             Ratio

Term                Effect  Effect     Coef  SE Coef  T-Value  P-Value   VIF

Constant                             0.3424   0.0481     7.12    0.001

Material           -0.9598  0.3830  -0.4799   0.0481    -9.99    0.000  1.00

InjPress           -0.1845  0.8315  -0.0922   0.0481    -1.92    0.113  1.00

InjTemp             0.0555  1.0571   0.0278   0.0481     0.58    0.589  1.00

CoolTemp           -0.1259  0.8817  -0.0629   0.0481    -1.31    0.247  1.00

Material*InjPress  -0.9918  0.3709  -0.4959   0.0481   -10.32    0.000  1.00

Material*InjTemp    0.1875  1.2062   0.0937   0.0481     1.95    0.109  1.00

Material*CoolTemp   0.0056  1.0056   0.0028   0.0481     0.06    0.956  1.00

InjPress*InjTemp   -0.0792  0.9239  -0.0396   0.0481    -0.82    0.448  1.00

InjPress*CoolTemp  -0.0900  0.9139  -0.0450   0.0481    -0.94    0.392  1.00

InjTemp*CoolTemp    0.0066  1.0066   0.0033   0.0481     0.07    0.948  1.00

Interpretation

For the insulation data, the manufacturer uses least squares regression to estimate the effects. These results are almost identical to the maximum likelihood results. The results indicate that:

·    The interaction between material and injection pressure has the greatest estimated effect (-0.9918) on the natural log of the standard deviation of insulation strength. The larger estimated effect does not indicate that the effect is statistically greater than the other effects.

·    Material has the second greatest effect (-0.9598) on the natural log of the standard deviation of insulation strength. The formula 2 material produced less variability in insulation strength than the formula 1 material.

You may want to refit the model excluding the nonsignificant terms.