In: Statistics and Probability
Statistics concepts for engineering management:
The data in the table provides: College GPA, High School GPA,
SAT total score, and number of letters of reference.
a. Generate a model for college GPA as a function of
the other three variables.
b. Is this model useful? Justify your
conclusion.
c. Are any of the variables not useful predictors?
Why?
CGPA HSGPA SAT REF
2.04 2.01 1070 5
2.56 3.4 1254 6
3.75 3.68 1466 6
1.1 1.54 706 4
3 3.32 1160 5
0.05 0.33 756 3
1.38 0.36 1058 2
1.5 1.97 1008 7
1.38 2.03 1104 4
4.01 2.05 1200 7
1.5 2.1 896 7
1.29 1.34 848 3
1.9 1.51 958 5
3.11 3.12 1246 6
1.92 2.14 1106 4
0.81 2.6 790 5
1.01 1.9 954 4
3.66 3.06 1500 6
2 1.6 1046 5
a.)
Below output was generated using regression from excel data analysis addin.
Model was built using CGPA as the dependent variable Vs HSGPA,SAT, REF as dependent variables.
Model coefficients are highlighted in green below. Model performance measures highlighted in yellow.
SUMMARY OUTPUT | ||||||
Regression Statistics | ||||||
Multiple R | 0.906865294 | |||||
R Square | 0.822404661 | |||||
Adjusted R Square | 0.786885593 | |||||
Standard Error | 0.501041207 | |||||
Observations | 19 | |||||
ANOVA | ||||||
df | SS | MS | F | Significance F | ||
Regression | 3 | 17.43781827 | 5.812606091 | 23.15389203 | 6.99009E-06 | |
Residual | 15 | 3.765634359 | 0.251042291 | |||
Total | 18 | 21.20345263 | ||||
Coefficients | Standard Error | t Stat | P-value | Lower 95% | Upper 95% | |
Intercept | -2.961593238 | 0.683200733 | -4.334880066 | 0.000589025 | -4.417801129 | -1.505385346 |
HSGPA | 0.026328666 | 0.203880871 | 0.1291375 | 0.898964517 | -0.408233124 | 0.460890456 |
SAT | 0.003714768 | 0.000740799 | 5.014542902 | 0.000153971 | 0.002135793 | 0.005293744 |
REF | 0.195979969 | 0.108310436 | 1.809428306 | 0.090465946 | -0.034878261 | 0.4268382 |
b.)
Yes model is useful since it performs well.
1. Adjusted R Square value of 0.787 seems reasonable.
2. Also the f-test is found to reject H0: regression model does not explain the variance at significance level .05. Highlighted in blue.
c.)
HSGPA is not useful for the model as suggested by the t-test result which shows p-value of 0.8989 which implies fail to reject H0: coefficient = 0 at .05 level of significance.
This is also supported by the plot of CGPA against HSGPA which shows little linear relationship between the variables.
Coefficients | Standard Error | t Stat | P-value | |
Intercept | -2.961593238 | 0.683200733 | -4.334880066 | 0.000589025 |
HSGPA | 0.026328666 | 0.203880871 | 0.1291375 | 0.898964517 |
SAT | 0.003714768 | 0.000740799 | 5.014542902 | 0.000153971 |
REF | 0.195979969 | 0.108310436 | 1.809428306 | 0.090465946 |