Anda di halaman 1dari 28

Correlation &

Multiple Regression
Analysis
Correlation Analysis
• Correlation indicates the direction, strength and
significance of the bivariate relationship among
variables that are included in the study

• It is the assessment of variation in one variable with


respect to the variation occurs in another variable

• It ranges between -1.0 to +1.0


Correlation Analysis
• Usually correlation coefficient greater than .60 is
considered as fairly strong correlation

• We need to consider whether the level of correlation


among variables is significant or not

• The correlation of .60 indicates that variance between


variables is ……………

• Most widely used method is Pearson Correlation


coefficient
Correlation Analysis
Job Perceived
Training Org Justice LMX Org Support
security Effectiveness

Training 1 .441* .508* .553* .419* .662*

Org Justice 1 .469* .601* .477* .574*

LMX 1 .588* .444* .413*

Org Support 1 .621* .531*

Job security 1 .472*

Perceived 1
Effectiveness

*. Correlation is significant at the 0.05 level


Multiple Regression Analysis
• Introduction:
• Statistical technique used to estimate the relationship
between a single dependent variable and two or more
independent variables

• It helps in understanding how the change occurs in the value


of dependent variable, when one or more independent
variables are changed
Multiple Regression Analysis
• Bank manager is interested to study the factors that
affects no. of credit cards used. The potential factors
identified were family size, family income and no. of
automobile used. For this, he collected data from
eight families in the town.
Multiple Regression Analysis
Family No. of Credit card Family Size Family Income ($000) No. of automobile
ID used (Y) (V1) (V2) used (V3)

1 4 2 14 1

2 6 2 16 2

3 6 4 14 2

4 7 4 17 1

5 8 5 18 3

6 7 5 21 2

7 8 6 17 1

8 10 6 25 2
Multiple Regression Analysis
• First, we consider correlation matrix

No. of
No. of Credit Family Size Family
Variables automobile
card used (Y) (V1) Income (V2)
owned (V3)

No. of credit cards used 1.000

Family Size .731 1.000

Family Income .614 .504 1.000

No. of automobile owned .206 .409 .458 1.000


Multiple Regression Analysis
• Concept of Multicollinearity:

• The multicollinearity refers to the correlation among


independent variables.

• In regression analysis, it is highly undesirable.

• The impact of multicollinearity is to reduce any single


IV’s predictive power by the extent it is associated
with other IVs.
Multiple Regression Analysis
• As multicollinearity increases, the unique variance
explained by each independent variable decreases

• Hence, the overall prediction increases much more


slowly as independent variables with high
multicollinearity are added

• To maximize the prediction from a fixed no. of IVs, we


should look for those IVs that have low multicollinearity
but high correlation with DV
Multiple Regression Analysis
• Since, the strongest correlation of dependent variable is with
family size. Therefore, we will consider it first as an
independent variable

• Regression equation becomes:

Y = b0 + b1V1 (general equation)

Y = 2.87 + .97V1
Multiple Regression Analysis
• Interpretation:

• b0 is an intercept. It has an interpretive value only when zero is a


conceptually valid value

• b1 is called regression coefficient. It is the estimated change in the


dependent variable for a unit change of the independent variable

• It should be statistically significant

• It indicates the extent to which the independent is associated with


the dependent variable
Multiple Regression Analysis

Model R R Square Adjusted R Square Std. Error of the


Estimate

1 .731 .534 .533 .414


Multiple Regression Analysis
• Coefficient of determination (R2):

• It represents the amount of variance in the DV explained by


the IV(s)

• It ranges from 1.0 (perfect prediction) to 0.0 (no prediction)

• It means that 53.4% variance in the no. of credit card used is


explained by variance in the family size
Multiple Regression Analysis
• Addition of 2nd IV:
• Task of the researcher is to expand the simple regression
model by adding more IVs that have the greatest additional
predictive power

• Now, include 2nd independent variable that has next highest


correlation with dependent variable (family income). Multiple
regression equation becomes:

Y = .482 + .63V1 + .216V2


Multiple Regression Analysis
Model R R Square Adjusted R Square Std. Error of the
Estimate

2 .802 .643 .622 .507

• Now R2 has increased to .643. It means that the 64.3%


variance in the no. of credit card used can be explained by the
variance in the family size and family income

• R2 has increased by a factor of 20.4% but Adjusted R2 has


increased 16.7%
Multiple Regression Analysis
• Adding 3rd IV:
• The only remaining IV is no. of automobiles owned. It has very
low correlation of .206 with DV

Model R R Square Adjusted R Square Std. Error of the


Estimate

3 .841 .707 .635 .528

Y = .286 + .635V1 + .200 V2 + .272V3


Multiple Regression Analysis
• R2 has increased to .707 (9.95 %) but Adjusted R2 is increased
only by 2.09 %

• This very slight increase in Adjusted R2 value is because of


two reasons:

• Weak correlation with DV


• Multicollinearity
Multiple Regression Analysis

Model R R Square Adjusted R Std. Error of the


Square Estimate

3 .841 .707 .635 .528

Sum of Squares df Mean Square F Sig.

Regression 15.684 3 5.228 4.306 .002

Residual 6.068 5 1.214

Total 21.752 8
Multiple Regression Analysis
Significance @
Unstandardized Standardized 95 %
T- Value VIF
Beta Coefficient Beta Coefficient confidence
level

(Constant) .286 3.716 .017

Family Size .635 .536 4.928 .004 1.620

Family Income .200 .403 3.991 .011 1.735

No. of automobile .272 .220 2.044 .102 3.179


owned
Multiple Regression Analysis
• Unstandardized Beta Coefficients

• IVs with larger regression coefficients would make a greater


contribution to the predicted value

• Standardized Beta Coefficient

• The variance in response scale and variability makes direct


interpretation problematic

• This problem is solved by the process of “Standardization”


Multiple Regression Analysis
• Standardization:
• It is a process through which statistical software converts all
variables to a common scale and variability

• In multiple regression table, there are two betas;


Unstandardized & Standardized beta.

• While using standardized beta value for measuring


comparative importance of IVs, two things should be
considered:
Multiple Regression Analysis
• First, standardized beta only guide for the relative importance
when multicollinearity is not very high (less than .65)

• These beta values can only be interpreted in the context of


other variables in the equation. If any other variable will be
added in the equation, all the relative measures will be
changed
Multiple Regression Analysis
• An industrial electric motors manufacturing firm is interested to
study the factors that affect clients’ perception (and hence sales)
about its motors. After thorough study, they find out following
factors are important in shaping customers’ perception:

1. Product Quality
2. Advertisement
3. Warranty & Claims
4. Price Flexibility
5. Complaint Resolution
Multiple Regression Analysis
• Firm collected data from 83 different customers who
are using its manufactured motors for at least three
years. Data was collected throughout country
Correlation
Product Warranty & Price Complaint Customers’
Advertisement
Quality Claims Flexibility Resolution Perception

Product Quality 1 .247* .310* -.340* .282* .587*

Advertisement 1 .382* .433* .401* .204*


Warranty &
1 .510* .684* .481*
Claims

Price Flexibility 1 .321* .502*

Complaint
1 .493*
Resolution

Customers’
1
Perception

*Correlation is significant at the 0.05 level


Model R R Square Adjusted R Square Std. Error of the
Estimate

1 .741a .549 .536 .393

Sum of df Mean F Sig.


Squares Square

Regression 47.860 5 9.572 18.820 .000b

Residual 39.673 78 .509

Total 87.533 83

a. Dependent Variable: Customers’ Perception

b. Predictors: (Constant), Product Quality, Advertisement, Warranty & Claims, Price Flexibility,
Complaint Resolution
Multiple Regression Table
Significance @
Unstandardized Standardized
T- Value 95 % confidence VIF
Beta Beta
level

(Constant) .435 2.946 .006

Product Quality .491 .438 4.828 .000 1.718

Advertisement .126 .118 1.194 .258 2.098

Warranty &
.202 .191 2.262 .059 2.375
Claims

Price Flexibility .342 .299 3.388 .002 1.791

Complaint
.366 .339 3.596 .001 2.101
Resolution

Anda mungkin juga menyukai