Chapter 08 – Cost Estimation
8-55 Cross-Sectional Regression; Ranks; Follows from 8-54 (35 min)
The regression results are shown below. Note that none of the
independent variables are significant.
To examine the possibility that the size of the company may be a
confounding factor in this analysis, an additional regression was run to
determine if the data for job openings and number of hires would be more
significant if normalized for the number of employees. For this analysis we
Also, note that the dependent variable in these regressions is a ranking
(1,2,3..), an ordinal number. The independent variables are real numbers
(any value along a continuum). The regression could be improved if we
could get a dependent variable that was not a ranking but a real number;
an example would be a rating, say, on a scale of zero to 100. This way we
would have more information about the relationships among the companies
being ranked.
8-61
Education.
Chapter 08 – Cost Estimation
8-55 (continued-1)
Regression results using normalized values for the independent variables.
This set of regressions shows there is very little relationship between
number of hires or number of job openings and the characteristics of the
Best Companies, as ranked by Fortune magazine. The criteria used by
the participants in the survey indicate their appreciation for other factors, as
noted in Fortune’s brief explanation of each ranking. The highest ranking
8-62
Education.
Chapter 08 – Cost Estimation
8-56 Cost Estimation; Regression Analysis (50 min)
1. The spreadsheet regression output for Plantcity is shown in
Exhibits 8-56A, B and C. Exhibit 8-56A shows the regression which
includes both predictors, sales dollars and sales units, while Exhibit
8-56B shows sales dollars only, and Exhibit 8-56C shows sales units
only.
Exhibit 8-56A (Units and Dollars)
8-63
Education.
Chapter 08 – Cost Estimation
Problem 8-56 (continued -1)
Exhibit 8-56B (Dollars)
Exhibit 8-56C (Units)
8-64
Education.
Chapter 08 – Cost Estimation
8-56 (continued -2)
The precision of the regression shown in 8-56A is good, with a
standard error of the estimate of 357 relative to a dependent variable
with values averaging about 3,000. Also, the reliability of the model is
The regression using sales dollars only (Exhibit 8-56B) is somewhat
worse (in terms of R-squared and F) while the regression on sales
units (Exhibit 8-56C) gives almost equivalent R-squared and standard
errors values to the model with both units and dollars. Because the
For further regression analysis on this data, consider the graphs
below which shows evidence of seasonality in the data.
8-65
Education.
Chapter 08 – Cost Estimation
8-56 (continued -3)
0 2 4 6 8 10 12
0
2
4
6
8
10
12
Dollars
Since the graphs show clear evidence of seasonality, another try of
the model with seasonality included would be a useful next step. The
8-66
Education.
Chapter 08 – Cost Estimation
8-56 (continued -4)
The seasonal model is shown in Exhibit 8-56D. Note the substantial
improvement in R-squared; also note that the seasonal variable is
significant. The coefficient on the seasonality variable is negative
because supplies expense does not rise as fast as units sold in
December.
Exhibit 8-56D
2. Predicted monthly figures for 2016 supplies expense using the
regression in Exhibit 8-56D:
8-67
Education.