STAT 2263 Linear Regression and Correlation

➢ Linear Regression

* Regression analysis is used to predict the value of one variable (the dependent variable) on
the basis of other variables (the independent variable).

* Dependent variable: usually denoted y

Independent variable: denoted x
Analysis between x and y in three ways:
o Correlation
o Determination
o Equation

➢ (1) Linear Correlation Coefficient, r

Correlation: The relationship between two variables (x and y)

r: measures the strength of the linear relationship between x and y.

-1 1

• Guideline for interpretation of r –1≤r≤1

|r|
0.00 No linear relationship
0.01 – 0.19 very weak
0.20 – 0.39 weak
0.40 – 0.69 modest “r” tells you only weak, strong, perfect
0.70 – 0.89 strong ___________________________________
0.90 – 0.99 very strong
1.00 perfect

MIDORI KOBAYASHI 1
STAT 2263 Linear Regression and Correlation

➢ (2) Coefficient of Determination, r2

❑ A measure of how much of the variation in dependent variable(y) can be "explained by"
the variation in independent variable (x).

For example,
x: pulse rate
y: white blood cell count and say r2 = 0.083

8.3% of the _____________________

variation in __________________________________________can
the white blood cell count be explained

by the ___________________
variation in ___________________________.
pulse rate The remaining __________%
91.7 is

unexplained (cannot be explained by pulse rate).

100-8.3

MIDORI KOBAYASHI 2
STAT 2263 Linear Regression and Correlation

➢ (3) Regression Equation

Given a collection of paired sample data, the regression equation,

yˆ b0 b1 x , algebraically describe the relationship between the two variables.

The graph of the regression equation is called the regression line or line of best fit.

RED

Interpretation of Slope:

As ______
x is increased by _______________________,
one unit ______
y is increased (or decreased) by

________units.
b
1
b
❖ The sign of ______ r
1 and ______must be the same.
positive
If b1 is positive, then r must be__________________.

negative
If b1 is negative, then r must be__________________.

Reference:
n(xy ) − (x)(y )
b1 ( slope) =
n ( x 2 ) − (  x ) 2
( y ) ( x )
b0 ( y − intercept ) = − b1  (n : # of pairs)
n n

MIDORI KOBAYASHI 3
STAT 2263 Linear Regression and Correlation

❖ Example (1)

A statistics professor claims that there is a distinct relationship between a student’s mark
and the student’s class attendance.

students A B C D E F G H

Stat Mark (%) 90 75 55 73 70 85 63 45

# of missed classes 2 7 13 9 7 4 13 16

(a) Determine the independent and dependent variables.

Does class attendance depend on the test mark?
Does the test mark depend on class attendance?

Independent variable: # of missed classes

Dependent variable: Stat Mark

(b) Is there a positive or negative correlation?

negative
_________________________ correlation

As the _____________________________increases,
# of missed classes

the stat mark ________________________.

decreases

(The graph is ____________________________).

right side down

The following is the output by SPSS

Model Summary and Parameter Estimates
Dependent Variable: StatMark
Model Summary Parameter Estimates
Equation R Square F df1 df2 Sig. Constant b1
Linear .942 96.735 1 6 .00006367 96.118 -2.999
The independent variable is ClassMissed.
(c) Write the regression equation using the output of SPSS.

yˆ b0 b1 x ^y = 96.118 - 2.999x

(d) Interpret the meaning of the regression coefficient, b1

b1(-2.999)
As the number of missed classes is increased by one class, the
student’s statistic mark (%) can be decreased by 2.999%
MIDORI KOBAYASHI 4

same unit
 STAT 2263 Linear Regression and Correlation

Model Summary and Parameter Estimates

Dependent Variable: StatMark
Model Summary Parameter Estimates
Equation R Square F df1 df2 Sig. Constant b1
Linear .942 96.735 1 6 .00006367 96.118 -2.999
The independent variable is ClassMissed.

(e) Find the coefficient of determination and interpret it.

r 2= 0.942
94.2% of the variation in stat marks can be
do not explained by the variation in the number of missed
round classes. The remaining 5.8% is unexplained.

100-94.2

(f) Find the correlation coefficient and interpret it.

r = r2 = 0.942 = 0.970566… =0.971 keep same

Since , r = -0.971 decimal place
b is negative
as r^2

very strong negative

There is a ____________________________________________________ linear relationship

# of missed classes
between _____________________________________________________ stat marks
and _____________________________________.

(g) Is there a linear relationship between the number of missed classes and the statistics marks at the 5%
significance level?

Since p-value = 0.00006367 < 0.05 , we reject H0.

There is a significant linear relationship between _______________________________________

the number of missed classes and

stat marks
__________________________________________________________ 5
at the _______% significance level.

Ref. table

MIDORI KOBAYASHI 5
STAT 2263 Linear Regression and Correlation

(h) Predict the stat marks for students who miss 3 classes and 20 classes.
Since there is a significant linear relationship and the correlation is _____________________________,
very strong we
can use the best fit equation to predict statistic marks within 2 – 16 missed classes.

ŷ = 96.118 – 2.999 x

When x =3

= 96.118 - 2.999 (3) = 87.121 = 87.1

When x = 20

The equation cannot be used to predict.

(20 is out of interval)

❖ Example (2) The following table shows the years of experience of 14 registered nurses and
their salaries.
Years of 0.5 2 4 5 7 9 10 12.5 13 16 18 20 22 25
Experience

Annual 45.2 48.3 53.6 46.8 54.3 68.5 58.3 54.6 72.6 76.2 72.6 94.3 75.3 82.3
Salary
(in 000’s)

(a) Determine the independent and dependent variables.

Independent variable: # of years of experience

Dependent variable: annual salary

(b) Is there a positive or negative correlation?

positive
_______________________ correlation

As the _____________________________________________
# of years of experience increases, the ______________________________
annual salary increases.
(The graph is ____________________________________.)
right side up
MIDORI KOBAYASHI 6
 STAT 2263 Linear Regression and Correlation

(c) Write the regression equation using the output from SPSS.

yˆ b0 b1 x = 44.326 + 1.722x

(d) Interpret the meaning of the regression coefficient, b1.

(1.722):
As the numbers of years of experience is increased by one year, the
annual salary of nurses will be increased by $1722 (1.722 thousands)

(e) Find the coefficient of determination and interpret it.

= 0.782

do not 78.2% of the variation in annual salaries of nurses can be

round explained by the variation in the number of years of experience.
The remaining 21.8% is unexplained
100-78.2

(f) Find the correlation coefficient and interpret it.

r = r2 = 0.782 = 0.88430 = 0.884

Since is positive , r = 0.884 (keep the same decimal places as r^2)

strong positive
There is a ____________________________________________________ linear relationship

between _____________________________________________________
years of experience and _____________________________________.
annual salaries

(g) Is there a linear relationship between the number of years of experience and the annual salary of
nurses at the 1% significance level?

MIDORI KOBAYASHI 7
STAT 2263 Linear Regression and Correlation

Since p-value = 0.0000271 < 0.01 , we reject H0.

p value a-value
There is a significant linear relationship between _______________________________________
the years of experiences and

__________________________________________________________
the salaries of nurses at the _______%
1 significance level.

Ref. table

P-value

(h) Predict the amount of salary if a nurse has 14 years of experience.

Years of 0.5 2 4 5 7 9 10 12.5 13 16 18 20 22 25

Experience

Annual 45.2 48.3 53.6 46.8 54.3 68.5 58.3 54.6 72.6 76.2 72.6 94.3 75.3 82.3
Salary
(in 000’s)

Since there is a significant linear relationship and the correlation is_____________________,

strong we can use the best
fit equation to predict the salaries within 0.5 – 25 years of experience.

yˆ 44.326 1.722x

substitute x=14 into the regression equation to get:

= 44.326 + 1.722 (14) = 68.434 = $68.434 (k) = $68434

MIDORI KOBAYASHI 8