Multiple Regression

Multiple Regression
 Multiple Regression Model
 Least Squares Method
 Multiple Coefficient of
 Determination
Model Assumptions
 Testing for Significance
 Using the Estimated Regression
Equation
for Estimation and Prediction
Multiple Regression Model
 Multiple Regression Model
The equation that describes how the dependent
variable y is related to the independent variables x1,
x2, . . . xp and an error term is:

y = b0 + b1x1 + b2x2 + . . . + bpxp + e

where:
b0, b1, b2, . . . , bp are the parameters, and
e is a random variable called the error term
 Multiple Regression Equation

 Multiple Regression Equation

The equation that describes how the
mean value of y is related to x1, x2, . . . xp
is:
E(y) = 0 + 1x1 + 2x2 + . . . + pxp
 Estimated Multiple Regression Equation

 Estimated Multiple Regression Equation

y =^b0 + b1x1 + b2x2 + . . . + bpxp

A simple random sample is used to compute

sample statistics b0, b1, b2, . . . , bp that are
used as the point estimators of the parameters
b 0, b 1, b 2, . . . , b p .
Estimation Process
Multiple Regression Model
Sample Data:
E(y) = 0 + 1x1 + 2x2 +. . .+ pxp + e x1 x 2 . . . x p y
Multiple Regression Equation . . . .
E(y) = 0 + 1x1 + 2x2 +. . .+ pxp . . . .
Unknown parameters are
b 0, b 1, b 2, . . . , b p

Estimated Multiple
Regression Equation
b0 , b1 , b2 , . . . , bp
provide estimates of yˆ b0  b1x1  b2x2  ...  bpxp
b 0, b 1, b 2, . . . , b p Sample statistics are
b0 , b1 , b2 , . . . , bp
Least Squares Method
• Least Squares Criterion

min  (yi  yˆi )2

 Computation of Coefficient Values

The formulas for the regression coefficients
b0, b1, b2, . . . bp involve the use of matrix algebra.
We will rely on computer software packages to
perform the calculations.
Multiple Regression Model

 Example: Programmer Salary Survey

A software firm collected data for a sample
of 20
computer programmers. A suggestion was
made that
regression analysis could be used to determine
if
The years
salary of experience,
was related score
to the years of on the aptitude
experience
test
and the
test,
scoreand corresponding
on the annual salary
firm’s programmer ($1000s)
aptitude test.
for a
sample of 20 programmers is shown on the next
slide.
Multiple Regression Model

Exper. Test Salary Exper. Test Salary

(Yrs.) Score ($000s) (Yrs.) Score ($000s)
4 78 24.0 9 88 38.0
7 100 43.0 2 73 26.6
1 86 23.7 10 75 36.2
5 82 34.3 5 81 31.6
8 86 35.8 6 74 29.0
10 84 38.0 8 87 34.0
0 75 22.2 4 79 30.1
1 80 23.1 6 94 33.9
6 83 30.0 3 70 28.2
6 91 33.0 3 89 30.0
Multiple Regression Model

Suppose we believe that salary (y) is

related to
the years of experience (x1) and the score on
the
programmer aptitude test (x2) by the
following
y = 0 + 1x1 + 2x2 + 
regression model:
where
y = annual salary ($000)
x1 = years of experience
x2 = score on programmer aptitude test
Solving for the Estimates of 0, 1, 2

Least Squares
Input Data Output
x1 x2 Computer b0 =
y Package b1 =
for Solving
4 78 b2 =
Multiple
24
Regression R2 =
7 100
43 Problems etc.
. .
.
. .
.
Solving for the Estimates of 0, 1, 2

 Excel’s Regression Equation Output

A B C D E
38
39 Coeffic. Std. Err. t Stat P-value
40 Intercept 3.17394 6.15607 0.5156 0.61279
41 Experience 1.4039 0.19857 7.0702 1.9E-06
42 Test Score 0.25089 0.07735 3.2433 0.00478
43
Estimated Regression Equation

SALARY
SALARY =
= 3.174
3.174 +
+ 1.404(EXPER)
1.404(EXPER) +
+ 0.251(SCORE)
0.251(SCORE)

Note: Predicted salary will be in thousands of dollars.

Interpreting the Coefficients

In multiple regression analysis, we

interpret each
regression coefficient as follows:
bi represents an estimate /Expected change
in y
corresponding to a 1-unit increase in xi when
all
other independent variables are held constant.
Interpreting the Coefficients

b
b11 =
= 1.404
1.404

Salary is expected to increase by $1,404 for

each additional year of experience (when the
variable
score on programmer attitude test is held
constant).
Interpreting the Coefficients

b
b22 =
= 0.251
0.251

Salary is expected to increase by $251 for

each
additional point scored on the programmer
aptitude
test (when the variable years of experience is
held
constant).
 Multiple Coefficient of Determination

 Relationship Among SST, SSR, SSE

SST = SSR +
SSE

 i
( y  y )2
= i
( ˆ
y  y )2
+  i i
( y  ˆ
y )2

where:
SST = total sum of squares
SSR = sum of squares due to regression
SSE = sum of squares due to error
Multiple Coefficient of Determination

 Excel’s ANOVA Output

A B C D E F
32
33 ANOVA
34 df SS MS F Significance F
35 Regression 2 500.3285 250.1643 42.76013 2.32774E-07
36 Residual 17 99.45697 5.85041
37 Total 19 599.7855
38
SSR
SST
Multiple Coefficient of Determination

R2 = SSR/SST

R2 = 500.3285/599.7855 = .83418
Adjusted Multiple Coefficient
of Determination

n 1
Ra2 2
 1  (1  R )
n  p 1

2 20  1
R 1  (1  .834179)
a  .814671
20  2  1
Assumptions About the Error Term 

The error  is
The error is aa random
random variable
variable with
with mean
mean of
of zero.
zero.

The
The variance of  ,, denoted
variance of by 
denoted by 22,, is
is the
the same
same for
for all
all
values
values of
of the
the independent
independent variables.
variables.

The
The values of  are
values of are independent.
independent.

The error  is
The error is aa normally
normally distributed
distributed random
random variable
variable
reflecting
reflecting the
the deviation
deviation between
between thethe yy value
value and and the the
expected
expected value
value ofof yy given by 00 +
given by + 11xx11 +
+ 22xx22 + + ppxxpp..
+ .. .. +
Testing for Significance

In
In simple
simple linear
linear regression,
regression, the
the FF and
and tt tests
tests provide
provide
the
the same
same conclusion.
conclusion.

In
In multiple
multiple regression,
regression, the
the FF and
and tt tests
tests have
have different
different
purposes.
purposes.
Testing for Significance: F Test

The
The FF test
test is
is used
used toto determine
determine whether
whether aa significant
significant
relationship
relationship exists
exists between
between the
the dependent
dependent variable
variable
and
and the
the set
set ofof all
all the
the independent
independent variables.
variables.

The
The FF test
test is
is referred
referred to
to as
as the
the test
test for
for overall
overall
significance.
significance.
Testing for Significance: t Test

If
If the
the FF test
test shows
shows an
an overall
overall significance,
significance, the
the tt test
test is
is
used
used toto determine
determine whether
whether each
each of
of the
the individual
individual
independent
independent variables
variables is
is significant.
significant.

A
A separate
separate tt test
test is
is conducted
conducted for
for each
each of
of the
the
independent
independent variables
variables in
in the
the model.
model.

We
We refer
refer to
to each
each of
of these
these tt tests
tests as
as aa test
test for
for individual
individual
significance.
significance.
Testing for Significance: F Test

Hypotheses H0: 1 = 2 = . . . = p = 0
Ha: One or more of the parameters
is not equal to zero.

Test Statistics F = MSR/MSE

Rejection Rule Reject H0 if p-value < a or if F > F ,

where F is based on an F distribution
with p d.f. in the numerator and
n - p - 1 d.f. in the denominator.
F Test for Overall Significance

Hypotheses H0: 1 = 2 = 0
Ha: One or both of the parameters
is not equal to zero.

Rejection Rule For  = .05 and d.f. = 2, 17; F.05 = 3.59

Reject H0 if p-value < .05 or F > 3.59
F Test for Overall Significance

 Excel’s ANOVA Output

A B C D E F
32
33 ANOVA
34 df SS MS F Significance F
35 Regression 2 500.3285 250.1643 42.76013 2.32774E-07
36 Residual 17 99.45697 5.85041
37 Total 19 599.7855
38
p-value used to test
for
overall significance
F Test for Overall Significance

Test Statistics F = MSR/MSE

= 250.16/5.85 = 42.76

Conclusion p-value < .05, so we can reject H0.

(Also, F = 42.76 > 3.59)
Testing for Significance: t Test

Hypotheses H 0 :  i 0
H a :  i 0

bi
Test Statistics t
sbi

Rejection Rule Reject H0 if p-value < a or

if t < -tor t > t where t
is based on a t distribution
with n - p - 1 degrees of freedom.
t Test for Significance
of Individual Parameters
Hypotheses H 0 :  i 0
H a :  i 0

For  = .05 and d.f. = 17, t.025 = 2.11

Rejection Rule
Reject H0 if p-value < .05, or
if t < -2.11 or t > 2.11
t Test for Significance
of Individual Parameters
 Excel’s Regression Equation Output
A B C D E
38
39 Coeffic. Std. Err. t Stat P-value
40 Intercept 3.17394 6.15607 0.5156 0.61279
41 Experience 1.4039 0.19857 7.0702 1.9E-06
42 Test Score 0.25089 0.07735 3.2433 0.00478
43

t statistic and p-value used to test for

the individual significance of
“Experience”
t Test for Significance
of Individual Parameters
 Excel’s Regression Equation Output
A B C D E
38
39 Coeffic. Std. Err. t Stat P-value
40 Intercept 3.17394 6.15607 0.5156 0.61279
41 Experience 1.4039 0.19857 7.0702 1.9E-06
42 Test Score 0.25089 0.07735 3.2433 0.00478
43

t statistic and p-value used to test for

the individual significance of “Test
Score”
t Test for Significance
of Individual Parameters
Test Statistics b1 1. 4039
 7. 07
sb1 . 1986
b2 . 25089
  3. 24
sb2 . 07735

Conclusions Reject both H0: 1 = 0 and H0: 2 = 0.

Both independent variables are
significant.
Testing for Significance: Multicollinearity

The
The term
term multicollinearity
multicollinearity refers
refers to
to the
the correlation
correlation
among
among the
the independent
independent variables.
variables.

When
When thethe independent
independent variables
variables are
are highly
highly correlated
correlated
(say,
(say, |r
|r || >
> .7),
.7), it
it is
is not
not possible
possible to
to determine
determine the
the
separate
separate effect
effect of
of any
any particular
particular independent
independent variable
variable
on
on the
the dependent
dependent variable.
variable.
Testing for Significance: Multicollinearity

If
If the
the estimated
estimated regression
regression equation
equation is
is to
to be
be used
used only
only
for
for predictive
predictive purposes,
purposes, multicollinearity
multicollinearity is
is usually
usually
not
not aa serious
serious problem.
problem.

Every
Every attempt
attempt should
should be
be made
made toto avoid
avoid including
including
independent
independent variables
variables that
that are
are highly
highly correlated.
correlated.
Using the Estimated Regression Equation
for Estimation and Prediction

The
The procedures
procedures for
for estimating
estimating the
the mean
mean value
value of
of yy
and
and predicting
predicting an
an individual
individual value
value of
of yy in
in multiple
multiple
regression
regression are
are similar
similar to
to those
those in
in simple
simple regression.
regression.

We
We substitute
substitute the
the given
given values
values of
of xx11,, xx22,, .. .. .. ,, xxpp into
into
the
the estimated
estimated regression
regression equation
equation andand use use the the
corresponding
corresponding value
value of
of yy as
as the
the point
point estimate.
estimate.
Using the Estimated Regression Equation
for Estimation and Prediction

The
The formulas
formulas required
required to
to develop
develop interval
interval estimates
estimates
for ^ y and for an individual value
for the
the mean
mean value
value of
of y and for an individual value
of
of yy are
are beyond
beyond the
the scope
scope ..

Software
Software packages
packages for
for multiple
multiple regression
regression will
will often
often
provide
provide these
these interval
interval estimates.
estimates.
Categorical Independent Variables

In
In many
many situations
situations we
we must
must work
work with
with categorical
categorical
independent
independent variables
variables such
such as
as gender
gender (male,
(male, female),
female),
method
method of
of payment
payment (cash,
(cash, check,
check, credit
credit card),
card), etc.
etc.

For
For example,
example, xx22 might
might represent
represent gender
gender where
where xx22 =
=00
indicates
indicates male
male and
and xx22 =
=11 indicates
indicates female.
female.

In
In this
this case,
case, xx22 is
is called
called aa dummy
dummy or
or indicator
indicator variable.
variable.
Categorical Independent Variables

 Example: Programmer Salary Survey

As an extension of the problem involving the
computer programmer salary survey, suppose
that
management also believes that the annual
salary is
related to whether the individual has a graduate
The years
degree of experience,
in computer the score
science on the
or information
systems. aptitude test, whether the individual has
programmer
a relevant graduate degree, and the annual salary
($000) for each of the sampled 20 programmers are
shown on the next slide.
Categorical Independent Variables

Exper. Test Salary Exper. Test Salary

(Yrs.) Score Degr. ($000s) (Yrs.) Score Degr. ($000s)
4 78 No 24.0 9 88 Yes 38.0
7 100 Yes 43.0 2 73 No 26.6
1 86 No 23.7 10 75 Yes 36.2
5 82 Yes 34.3 5 81 No 31.6
8 86 Yes 35.8 6 74 No 29.0
10 84 Yes 38.0 8 87 Yes 34.0
0 75 No 22.2 4 79 No 30.1
1 80 No 23.1 6 94 Yes 33.9
6 83 No 30.0 3 70 No 28.2
6 91 Yes 33.0 3 89 No 30.0
Estimated Regression Equation

y =^ b0 + b1x1 + b2x2 + b3x3

where:
y =^annual salary ($1000)
x1 = years of experience
x2 = score on programmer aptitude test
x3 = 0 if individual does not have a graduate degree
1 if individual does have a graduate degree

x3 is a dummy variable
Categorical Independent Variables

 Excel’s Regression Statistics

A B C
23
24 SUMMARY OUTPUT
25
26 Regression Statistics
27 Multiple R 0.920215239
28 R Square 0.846796085
29 Adjusted R Square 0.818070351
30 Standard Error 2.396475101
31 Observations 20
32
Categorical Independent Variables

 Excel’s ANOVA Output

A B C D E F
32
33 ANOVA
34 df SS MS F Significance F
35 Regression 3 507.896 169.2987 29.47866 9.41675E-07
36 Residual 16 91.88949 5.743093
37 Total 19 599.7855
38
Categorical Independent Variables

 Excel’s Regression Equation

Output A B C D E
38
39 Coeffic. Std. Err. t Stat P-value
40 Intercept 7.94485 7.3808 1.0764 0.2977
41 Experience 1.14758 0.2976 3.8561 0.0014
42 Test Score 0.19694 0.0899 2.1905 0.04364
43 Grad. Degr. 2.28042 1.98661 1.1479 0.26789
44

Not significant
 Categorical Independent Variables

 Excel’s Regression Equation

Output
A B F G H I
38
39 Coeffic. Low. 95% Up. 95% Low. 95.0% Up. 95.0%
40 Intercept 7.94485 -7.701739 23.5914 -7.7017385 23.591436
41 Experience 1.14758 0.516695 1.77847 0.51669483 1.7784686
42 Test Score 0.19694 0.00635 0.38752 0.00634964 0.3875243
43 Grad. Degr. 2.28042 -1.931002 6.49185 -1.9310017 6.4918494
44
More Complex Categorical Variables

If
If aa categorical
categorical variable
variable has
has kk levels,
levels, kk -- 1
1 dummy
dummy
variables
variables are
are required,
required, with
with each
each dummy
dummy variable
variable
being
being coded
coded as
as 00 or
or 1.
1.

For
For example,
example, aa variable
variable with
with levels
levels A,
A, B,
B, and
and CC could
could
be
be represented
represented byby xx11 and
and xx22 values
values of
of (0,
(0, 0)
0) for
for A,
A, (1,
(1, 0)
0)
for
for B,
B, and
and (0,1)
(0,1) for
for C.
C.

Care
Care must
must be
be taken
taken in
in defining
defining and
and interpreting
interpreting the
the
dummy
dummy variables.
variables.
More Complex Categorical Variables

For example, a variable indicating level of

education could be represented by x1 and x2
values as follows:

Highest
Degree x1 x2
Bachelor’s 0 0
Master’s 1 0
Ph.D. 0 1
Residual Analysis

 For simple linear regression the residual plot

against
ŷ
and the residual plot against x provide the
 same information.
In multiple regression analysis it is preferable
to use the residual plotŷ against to determine
if the model assumptions are satisfied.
Standardized Residual Plot Against ŷ

 Standardized residuals are frequently used in

residual plots for purposes of:
• Identifying outliers (typically, standardized
residuals < -2 or > +2)
• Providing insight about the assumption that
the error term e has a normal distribution
 The computation of the standardized residuals
in multiple regression analysis is too complex
to be done by hand
 Excel’s /SAS/e-views Regression tool
can be used
Standardized Residual Plot Against ŷ

 Excel Value Worksheet

A B C D
28
29 RESIDUAL OUTPUT
30
31 Observation Predicted Y Residuals Standard Residuals
32 1 27.89626052 -3.89626052 -1.771706896
33 2 37.95204323 5.047956775 2.295406016
34 3 26.02901122 -2.32901122 -1.059047572
35 4 32.11201403 2.187985973 0.994920596
36 5 36.34250715 -0.54250715 -0.246688757
Standardized Residual Plot Against ŷ
Outlier

Standardized Residual Plot

3

2
Residuals
Standard
1

0
0 10 20 30 40 50
-1

-2

-3
Predicted Salary
Logistic Regression

 In many ways logistic regression is like

ordinary regression. It requires a dependent
variable, y, and one or more independent
 variables.
Logistic regression can be used to model
situations in which the dependent variable, y,
may only assume two discrete values, such as
 0 and
The 1.
ordinary multiple regression model is not
applicable.
 Logistic Regression

 Logistic Regression Equation

The relationship between E(y) and x1, x2, . . . ,
xp is
better described by the following nonlinear
 0   1x1  2x2   pxp
equation. e
E(y)   0   1x1  2x2   pxp
1 e
 Logistic Regression

 Interpretation of E(y) as a
Probability in Logistic Regression
If the two values of y are coded as 0 or 1,
the value
of E(y) provides the probability that y = 1 given
a
particular
E(y) set of values
estimate of Pfor
(y x11|, xx12,,x.2 ,.. ,,xxp )p.
 Logistic Regression

 Estimated Logistic Regression Equation

b0 b1x1b2x2 bpxp
e
yˆ  b0 b1x1b2x2 bpxp
1 e

A simple random sample is used to

compute sample statistics b0, b1, b2, . . . , bp
that are used as the point estimators of the
parameters b0, b1, b2, . . . , bp.
Logistic Regression

 Example: Simmons
Stores
Simmons’ catalogs are expensive and
Simmons
would like to send them to only those customers
who
have the highest probability of making a $200
Simmons’ management thinks that annual
purchase
spending
using the discount coupon included in the
at catalog.
Simmons Stores and whether a customer has
a
Simmons credit card are two variables that might
be
helpful in predicting whether a customer who
receives
the catalog will use the coupon to make a $200
Logistic Regression

 Example: Simmons
Stores
Simmons conducted a study by sending out
100
catalogs, 50 to customers who have a Simmons
credit
card and 50 to customers who do not have the
card.
At the end of the test period, Simmons noted for
each of
the 100 customers:
1) the amount the customer spent last year at
Simmons,
2) whether the customer had a Simmons credit
card, and
3) whether the customer made a $200 purchase.
Logistic Regression
x1 x2 y
 Simmons Test Data (partial)
Annual Spending Simmons $200
Customer ($1000) Credit Card Purchase

1 2.291 1 0
2 3.215 1 0
3 2.135 1 0
4 3.924 0 0
5 2.528 1 0
6 2.473 0 1
7 2.384 0 0
8 7.076 0 0
9 1.182 1 1
10 3.345 0 0
 Logistic Regression

 Simmons Logistic Regression Table (using Minitab)

Odds 95% CI
Predictor Coef SE Coef Z p Ratio Lower Upper

Constant -2.1464 0.5772 -3.72 0.000

Spending 0.3416 0.1287 2.66 0.008 1.41 1.09 1.81
Card 1.0987 0.4447 2.47 0.013 3.00 1.25 7.17

Log-Likelihood = -60.487
Test that all slopes are zero: G = 13.628, DF = 2, P-Value = 0.001
 Logistic Regression

 Simmons Estimated Logistic Regression Equation

 2.14640.3416x11.0987x2
e
yˆ   2.14640.3416x11.0987x2
1 e
 Logistic Regression

 Using the Estimated Logistic Regression Equation

• For customers that spend $2000 annually
and do not have a Simmons credit card:
e 2.14640.3416(2)1.0987(0)
yˆ   2.14640.3416(2)1.0987(0)
0.1880
1 e
• For customers that spend $2000 annually
and do have a Simmons credit card:
e 2.14640.3416(2)1.0987(1)
yˆ   2.14640.3416(2)1.0987(1)
0.4099
1 e
 Logistic Regression

 Testing for Significance

Hypotheses H0: 1 = 2 = 0
Ha: One or both of the parameters
is not equal to zero.
Test Statisticsz = bi/sbi

Rejection Rule Reject H0 if p-value < a

 Logistic Regression

 Testing for Significance

Conclusions For independent variable x1:

z = 2.66 and the p-value = .008.
Hence, b1 = 0. In other words,
x1 is statistically significant.

For independent variable x2:

z = 2.47 and the p-value = .013.
Hence, b2 = 0. In other words,
x2 is also statistically significant.
 Logistic Regression

 Odds in Favor of an Event Occurring

P(y 1| x1 , x2 , , xp ) P(y 1| x1 , x2 , , xp )

odds  
P(y 0| x1 , x2 , , xp ) 1  P(y 1| x1 , x2 , , xp )

 Odds Ratio

odds1
Odds Ratio 
odds0
 Logistic Regression

 Estimated Probabilities

Annual Spending
$1000 $2000 $3000 $4000 $5000 $6000 $7000

0.3305 0.4099 0.4943 0.5791 0.6594 0.7315 0.7931

Yes
Credit
Card 0.1413 0.1880 0.2457 0.3144 0.3922 0.4759 0.5610
No

Computed
earlier
Logistic Regression

 Comparing Odds
Suppose we want to compare the odds of
making a
$200 purchase for customers who spend $2000
annually
and have a Simmons credit card to the odds of
making a
$200 purchase .4099
estimatefor customers
of odds1  who spend $2000
.6946
annually 1- .4099
and do not have a Simmons .1880
credit card.
estimate of odds0  .2315
1- .1880
.6946
Estimate of odds ratio  3.00
.2315