Question 1: fit the model
##
## Call:
## lm(formula = logY ~ logX1 + logX2 + logX3, data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.3987 -0.1264 0.0138 0.1118 0.2937
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 3.8646 2.2334 1.730 0.097 .
## logX1 4.9504 0.2557 19.363 9.78e-16 ***
## logX2 -5.6537 0.3858 -14.656 3.72e-13 ***
## logX3 -3.5030 0.3858 -9.081 4.56e-09 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.1827 on 23 degrees of freedom
## Multiple R-squared: 0.9669, Adjusted R-squared: 0.9626
## F-statistic: 224.1 on 3 and 23 DF, p-value: < 2.2e-16
Question 2: full model with the quadratic and interaction terms
##
## Call:
## lm(formula = logY ~ logX1 + logX2 + logX3 + logX2_and_logX3 +
## logX1_and_logX3 + logX1_and_logX2 + logX1_quadratic + logX2_quadratic +
## logX3_quadratic, data = data)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.304034 -0.112395 -0.005587 0.116521 0.266156
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -239.1097 157.2494 -1.521 0.147
## logX1 29.4930 34.6778 0.850 0.407
## logX2 16.8684 36.9321 0.457 0.654
## logX3 74.6512 52.3750 1.425 0.172
## logX2_and_logX3 -1.6384 4.4973 -0.364 0.720
## logX1_and_logX3 -3.4905 2.9806 -1.171 0.258
## logX1_and_logX2 -2.1849 2.9806 -0.733 0.474
## logX1_quadratic -0.5695 2.8232 -0.202 0.843
## logX2_quadratic -0.8791 6.3898 -0.138 0.892
## logX3_quadratic -7.1965 6.3898 -1.126 0.276
##
## Residual standard error: 0.1941 on 17 degrees of freedom
## Multiple R-squared: 0.9724, Adjusted R-squared: 0.9578
## F-statistic: 66.52 on 9 and 17 DF, p-value: 1.776e-11
Question 3: ANOVA table of the fullModel
| Df | Sum Sq | Mean Sq | F value | Pr(>F) | |
|---|---|---|---|---|---|
| logX1 | 1 | 12.51 | 12.51 | 332.04 | 0.0000 |
| logX2 | 1 | 7.17 | 7.17 | 190.23 | 0.0000 |
| logX3 | 1 | 2.75 | 2.75 | 73.03 | 0.0000 |
| logX2_and_logX3 | 1 | 0.01 | 0.01 | 0.13 | 0.7201 |
| logX1_and_logX3 | 1 | 0.05 | 0.05 | 1.37 | 0.2577 |
| logX1_and_logX2 | 1 | 0.02 | 0.02 | 0.54 | 0.4735 |
| logX1_quadratic | 1 | 0.00 | 0.00 | 0.04 | 0.8425 |
| logX2_quadratic | 1 | 0.00 | 0.00 | 0.02 | 0.8922 |
| logX3_quadratic | 1 | 0.05 | 0.05 | 1.27 | 0.2757 |
| Residuals | 17 | 0.64 | 0.04 |
Question 4: residual VS. fitted value
Question 5: Box-Cox test for response variable
From the plot above, we can see that the 95% interval of lambda contains 0, and the upper bond and lower bond are very close to 0. So the value of lambda should be approximately 0, which suggests that the natural log transformation should be taken for the respone variable.
Question 6: Box-Tidwell test for x1
## Score Statistic p-value MLE of lambda
## -0.3145565 0.7530984 -0.2304977
##
## iterations = 2
From the result above, we can see the p-value is 0.75 > 0.05, which suggests no transformation needed.
Question 7: two-way tables
Two-way table of Load and Amplitude of Loading Cycle| low | 0 | high | mean | |
|---|---|---|---|---|
| low | 7.32 | 6.70 | 6.09 | 6.70 |
| 0 | 7.02 | 6.33 | 5.79 | 6.38 |
| high | 6.58 | 5.92 | 5.26 | 5.92 |
| mean | 6.97 | 6.32 | 5.71 | 6.33 |
| low | 0 | high | mean | |
|---|---|---|---|---|
| low | 5.82 | 6.76 | 7.53 | 6.70 |
| 0 | 5.42 | 6.44 | 7.28 | 6.38 |
| high | 5.17 | 5.98 | 6.61 | 5.92 |
| mean | 5.47 | 6.39 | 7.14 | 6.33 |
| low | 0 | high | mean | |
|---|---|---|---|---|
| low | 6.03 | 6.93 | 7.96 | 6.97 |
| 0 | 5.58 | 6.48 | 6.89 | 6.32 |
| high | 4.80 | 5.76 | 6.57 | 5.71 |
| mean | 5.47 | 6.39 | 7.14 | 6.33 |