Data Models and Decisions
Autor: Adnan • April 17, 2018 • 1,230 Words (5 Pages) • 719 Views
...
(Thus, for example, Region 11 had 54,061 portable phones and its peak demand was 6,878 circuits.) David used linear regression for this purpose. The resulting computer output is shown below:
SUMMARY
OUTPUT
[pic 2]
Regression Statistics
Multiple R
0.97533
R Square
0.95127
Adjusted R Square
0.94856
Standard Error
3381.29
Observations
20
ANOVA
df
SS
MS
F
Regression
1
4,017,125,436
4,017,125,436
351.358278
Residual
18
205,796,369
11,433,131
Total
19
4,222,921,805
Coefficients
Standard Error
t Stat
P-value
Intercept
-9,159.91
1,396.64
-6.55855 3.6647E-06
Number of Phones
0.336564
0.017955
18.74455
2.9406E-13
---------------------------------------------------------------
15.063 Summer 2003 Final Examination p.5 of 7
- (5 points) Write a complete equation for the simple linear regression model that incorporates the estimated coefficients provided by this computer output. Make sure to define in words all the variables used in this equation.
- (5 points) Interpret, in managerial terms, the meaning of each of the two regression coefficients (including the units in which each is expressed).
David also produced a histogram of the regression residuals, and a plot of the residuals against number of portable phones, as shown below.
Frequency
---------------------------------------------------------------
6
5
4
3
2
1
0
---------------------------------------------------------------
Histogram
Residuals
10000
8000
6000
Residual
4000
2000
0
0
50,000
100,000
150,000
200,000
-2000
[pic 3]
-4000
-6000
-5000 -3000 -1000 1000 3000 5000 7000 More
Number of Phones
Residual Values
For each of the statements (A) to (E) below, indicate, by circling the correct answer, whether you think the statement is true or false. If you answered “TRUE”, please provide a brief justification of your answer in the space provided. (No justification is needed if you answered “FALSE”).
(c) (5 points) These graphs show evidence of:
1. non-Normal noise.
TRUE - FALSE
If “TRUE”, how do you know?
2. autocorrelation.
TRUE - FALSE
If “TRUE”, how do you know?
3. overspecification.
TRUE - FALSE
If “TRUE”, how do you know?
4. multicollinearity.
TRUE - FALSE
If “TRUE”, how do you know?
5. heteroscedasticity.
TRUE - FALSE
If “TRUE”, how do you know?
(d) (5 points) Explain how you would go about correcting the most serious problem of the ones you found in part (c). If applicable, how would you modify the data and/or the model to implement this change?
...