1

THE LEAST SQUARES LINE (other names ‘Best-Fit Line’ or ‘Regression Line’)

Problem:
A sales manager noticed that the annual sales of his employees increase with years of experience. To estimate the annual sales for his
potential new sales person he collected data concerning annual sales and years of experience of his current employees: We’ll use his data to
create a formula that will help him estimate annual sales based on years of experience.

Years of experience 1 3 8 10 13
150
Annual sales in thousands 80 100 110 120 140 140
130
120

Sales
110
100
Work: 90
80
Figure to the right shows scatter graph of his data. Each point represents data on
70
one single current employee. For example: the first employee has 1 year of
60
experience and made 80 thousands in sales, so he is represented by the point with
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
coordinates (1,80).
Years

We’ll create equation of the least squares line, which is also called best-fit
line or regression line. The line is passing in between our points while the
sum of the squares of the vertical distances from the data points to the line
is as small as possible. The picture to the left shows the least square line
passing in between our points, and the distances d1, d2, d3, d4, and d5. The
equation of the least square line is found by minimizing the sum:
(d1 ) 2  (d 2 ) 2  (d 3 ) 2  (d 4 ) 2  (d 5 ) 2
The procedure will be omitted in this paper. The final result of the
minimization are formulas that let us calculate coefficients a and b for
equation of the least square line y  ax  b . This equation may be used for
the prediction of sales.

FORMULAS FOR COEFFICIENTS OF THE LEAST SQUARES LINE:

n xy   x  y   y  a   x

a b

n  x   x 
2
 2
n

In this formula n stands for number of observed cases. In our case that is 5 employees. So: n5
Symbol  is called “sigma” and stands for the “sum of all”. In our case:

 x  1  3  8  10  13  35 (sum of all x values)

 y  80  100  110  110  120  140  550 (sum of all y values)

x 2
 12  32  82  10 2  132  343 (sum of squares of x values)

 xy  1 80  3 100  8 110  10 120  13 140  4280 (sum of xy products)

 2
We are now ready to use results of our calculations in the formula:

5  4280  35  550 2150 550  4.388  35 396.42

a   4.388 b   79.284
5  343  35 2 490 5 5

That means that the equation of the least square line is y  4.388 x  79.284

Conclusion:
We created the formula y  4.388 x  79.284 that may be used to predict sales in thousands for a future employee. For example, formula
may predict that an employee with x=15 years of experience will generate:
y  4.388  15  79.284  145 thousands in sales per year.

Is our prediction reliable?

Once an equation is found for the least square line, we need to have some way of judging just how good the equation is for predictive
purposes. In order to have a quantitative basis for confidence in our predictions, we need to calculate coefficient of correlation, denoted r. It
may be calculated using the following formula:

FORMULA FOR COEFFICIENT OF CORRELATION

n xy   x  y 

r
  2
 
n  x 2   x   n  y 2   y 
2

We’ll calculate the coefficient of correlation for data in our example:

y  80 2  100 2  110 2  120 2  140 2  62500

2

5  4280  35  550 2150 2150

r    0.971
5  343  35  5  62500  550
2 2
490  10000 2213.594

The value of r that is close to 1 indicates that our formula will give us a reliable prediction for sales level, based on years of experience of the
employee.

What does coefficient of correlation tell us?

The correlation coefficient is always a number between  1 and 1 . The picture bellow shows how its value numerically describes our data:

The equation may be used as a source for reliable prediction if the correlation coefficient is a number that is close to  1 or 1 . That means
that your observed values are close to the least square line (second and fourth picture above).
If not, the value of the correlation coefficient is closer to 0. Such a small value for the coefficient of correlation indicates that the observed
data are widely spread around, so our formula is not reliable source of prediction (like on first and third picture above).
It is usually more convenient to numerically measure reliability of our formula using the square of correlation coefficient. Some books and
computer software are using symbol R 2 for it (read as ‘r square’) In our example;

R 2  0.9712  0.943
R square is always a number between 0 and 1. Values of R 2 that are close to 1 indicate reliable formula.
 3
Problem:
A sales manager noticed that the annual sales of his employees increase with years of experience. We’ll use Excel to graph his data and create
a formula that will help him estimate annual sales based on years of experience.

Years of experience 1 3 8 10 13
Annual sales in thousands 80 100 110 120 140

The final picture shown below features the equation of the least squares line y  4.3878 x  79.286 and the value of R square R 2  0.9434

Practice questions:
1) Find the equation of the least squares line and the coefficient of correlation for the data shown:
x 3 5 7 8
y 8.7 7.9 6.2 5.8

2) The table to the right shows the relationship between MPG (miles per gallon) and HP x=HP y=MPG
(horsepower) for some sports cars produced in year 2015. Mitcubishi Lancer 148 29
Find the equation of the least squares line and use it to estimate MPG for Chevrolet Corvette that
Mazda Miata 158 23
has 460 HP engine.
Subaru WRX 268 24
Porche Boxter 315 23
Chevy Camaro 323 20
Ferrari Spider 570 14
3) Millers are buying a house in NWI so they’d like to estimate yearly house taxes in Valparaiso. Using data from ReMax web site they
obtained 2014 taxes for 5 houses in that city. Use the table to create the least squares line equation and its coefficient of correlation. Use this
equation to estimate yearly taxes for a $200,000 house in Valparaiso.

x=house price in thousands 425 340 140 250 70

y=tax paid in year 2014 3570 2370 1000 1700 810
 4
4) The table below gives monthly sales y (in thousands), corresponding to advertising expenditures x (in thousands).

Advertising expenditures x 0 1 2 3 4 5 6
Monthly sales y 3 9 11 15 17 20 27

Create the least squares line. Examine if advertising expenditures appear to have strong effect on monthly sales by calculating and
interpreting R 2 , and if so, predict the monthly sales if 11 thousand dollars is spent on advertising.

5) A hospital conducted study to determine relation between age and blood pressure of their patients. The table bellow shows collected data.
Find the equation of the least squares line and use it to calculate blood pressure of a 50 years old patient.

Age x 43 48 56 61 67 70
Pressure y 128 120 135 143 141 152

6) Find the equation of the least squares line for the data shown below:
x 0 5 10 20
y 8 7 5 2

7) A researcher wishes to see whether there is a relationship between number of hours of study and test scores on exam, so she collected
data shown in the table bellow. Find the equation of the least squares line and use it to calculate how many hours should a student study to
obtain 93 percent on the test.
Hours of study x 6 2 1 5 2 3
Score on test y 82 63 57 88 68 75

8) The table bellow compares rents for one-bedroom and two-bedroom apartments in 7 different cities. Find the equation of the least
squares line and R square. A one-bedroom rent in a doorman building in Lower Manhattan averaged $3000 in august 2007. Calculate a two
bedroom rent using the obtained formula.

One-bedroom rent x 782 486 451 529 618 520 845

Two bedroom rent y 1223 902 739 954 1055 875 1455

9) Thanks to the progress of science the cancer survival rate is improving over time. The table shows 10-year survival rate for ovary cancer
in England, diagnosed in year 1971, 1981, 1991 and 2001. According to the table 27% of the patients diagnosed in year 2001 survived over
10 years. Find the equation of the least squares line and R square. Use it to estimate the 10-year ovary cancer survival rate for the patients
who are diagnosed in year 2015.

diagnosis year 71 81 91 101

percent lived over 10 years 18 22 25 27

10) Millers are buying a house in NWI so they’d like to estimate yearly house taxes in Crown Point. Using data from ReMax web site they
obtained 2014 taxes for 6 houses in that city. Use the table to create the least squares line equation. Use this equation to estimate yearly taxes
for a $200,000 house in Crown Point.

x=price in thousands 155 125 205 275 110 780

y=tax for year 2014 3096 1217 2819 3852 877 13390