Sunteți pe pagina 1din 9

# SAM

## Case: Estimating Checking Account Balance

WAC

17010057
17010057
WAC-1
SAM

1-Developing Model

Basic Model
Table 1 and Table 2 show the statistical evaluation of the basic model. Almost 60% of the variation
is explained by the independent variable used in this basic model. The omission of statistically
insignificant variables (Debit and Interest as shown in Table 3) cannot be explained intuitively,
therefore improved model is developed. Equation for basic model is given below.
Bal= 186.959+85.165 ATM + 65.017 S +413.527R +54.99L +27.463F+E
Where
Bal= Balance I = Interest, S = Services, L= Lahore, R = Rawalpindi, F = Faisalabad, E=Error

Table 1

Table 2

Page 1 of 7
17010057
WAC-1
SAM

Table 3
Improved Model
Improved model is developed by transforming the dependent variable into Natural Logarithm and
by taking the square of one of independent variable i.e. Services. Squaring Services is explained
by the jump in adjusted R square (from 0.485 to 0.560).
Figure A and Table 4 explains the Jump in adjusted R square. Figure A represents the model
summary for model in which services square is not considered. When service square is used in the
regression there is nice jump in the adjusted R square.
Though the adjusted R square is reduced to 0.560 as compared to basic model but this improved
model includes Interest as an independent variable which is gives economic sense. Addition of
interest in model also reduces the standard error of estimates to 0.49689.
The model equation is given as below.
Ln (Bal) = 4.796 +0.074 ATM + 0.531S 0.044 S^2 +2.961 I + 0.368L + 0.384 R 0.088F+E
Where
Bal= Balance I = Interest, S = Services, L= Lahore, R = Rawalpindi, F = Faisalabad, E=Error

Figure A

Page 2 of 7
17010057
WAC-1
SAM

Table 4

Table 5

2-Independent Variables

In the improved model Debit Card is not statistically significant. Table 6 shows that after addition
of debit in the regression model the adjusted R square just jumped from 0.560 to 0.564. This
increase of 0.04 is very minimal. Another important change observed in the block diagram in Table
7 is that none of the dummy variable remain statistically significant after addition of debit in the
regression model. Addition of debit card does not significantly enhance the variation explained in
dependent variable and is not statistically significant, hence it is dropped from the model. Table 6
and Table 7 show that there is a good jump in adjusted R square with addition of new variable
expect for debit card.
Cognitively, from the data provided, all the customers who do not have a debit card have still used
ATM.

Page 3 of 7
17010057
WAC-1
SAM

Table 6

Table 7

## 3-Statistical Significance of Variables

Row 4 of Table 8 shows that model as a whole is statistically significant as the P-Value is less than
0.05.
Table 4 provides the information regarding individual regression coefficients. Individual
Coefficients are statistically significant, the stated P-values are less than 0.05. In case of dummy
variable of city, if any of the dummy variable is statistically significant than we should consider

Page 4 of 7
17010057
WAC-1
SAM

all dummy variables in our regression. In this approved model, P-Value of Rawalpindi is 0.043
which is less than 0.05 so all dummy variables are statistically significant.

Table 8

## 4-Variation Explained by Model

Table 4 shows that adjusted R square equals to 0.560. This explains that a total of 56% variation
in dependent variable is explained by variation in independent variables.

5-Checking Assumptions

## Assumption 1: Residuals follow normal probability distribution.

Figure B shows that the residuals follow the normal probability distribution.
Assumption 2: Variation of residual along regression equation is same.
Figure C does not show heteroscedasticity, therefore further investigation is not required.
Assumption 3: The residuals are independent.
From Table 9, Models Durbin Watson is 1.765 which lies in the inconclusive rage (1.3349-
1.8505), but it is very Close to the acceptance range, therefore in this case Null Hypothesis i.e.
there is no autocorrelation is accepted.
Assumption 4: Multicollinearity
Table 5 shows the variance inflation factor. The calculated values are in acceptable limits. Though
service and service square VIF values are greater than 10, but mathematically one quantity is
square of other so this is acceptable.
Assumption 5: Linear relation

Page 5 of 7
17010057
WAC-1
SAM

Figure D and Figure E show the relationship between Y and X variable that is linear. In this model

Figure B Figure C

Figure D Figure E

Table 9

Page 6 of 7
17010057
WAC-1
SAM

## 6-Interperation of Regression Equation

Ln (Bal) = 4.796 +0.074 ATM + 0.531S 0.044 S^2 +2.961 I + 0.368L + 0.384 R 0.088F+E
The signs of coefficients with independent variables (ATM, Interest) have expected signs.
Variables included in the regression equation are all significant. These signs imply that these
variables have direct relationship with balance. If any of these variable will increase the balance
will increase and if any of these variable will decrease the balance will decrease. For example if
ATM transactions will increase one unit there will be 7.4% increase in balance holding other
independent variable fixed. If interest will by 1 unit increase there will be 296% increase in
balance.
Keeping other variables fixed, balance with respect to services is increasing at a decreasing rate
till it reaches the inflation point. The maximum balance achieved is at inflation point i.e. 6.
Assuming that all other independent variables are fixed, the balance of a customer in Lahore will
be 36.8% higher relative to customers balance in Sialkot. Similarly the balance of a customer in
Faisalabad will be 8.8% less than the balance of a customer in Sialkot. A customer in Rawalpindi
will have 38.4% more balance than a customer in Sialkot

7-Estimating Balance

Ln (Bal) = 4.796 +0.074 ATM + 0.531S 0.044 S^2 +2.961 I + 0.368L + 0.384 R 0.088F+E
1. 793.2
2. 954.5
3. 792.6

## 8-Insight for Social Bank

Customers are encouraged to keep more balance in the Social Bank if they are getting more
interest. Balance has a strong direct relation with the interest. With every unit increase in interest
rate the balance is increased by 296%, which is a huge increment. Therefore to increase the bank
balance Social Bank should offer more interest. Of course this cannot be increased beyond certain
limit or the bank would incur loses.
ATM transactions do increase the balance but just by 7.4%. Social Bank can tap this option too
but the effect will be very minimal as compared to interest rate. Furthermore Services provided by
the bank encourages the customer to keep more balance but this increase is at a decreasing rate i.e.
after reaching the inflation point the balance will start to decrease.

Page 7 of 7
17010057
WAC-1
SAM

A maximum of 6 services should be provided by Social Bank to any customer, after which the
balance will start to decrease. Customers in Rawalpindi tend to have highest balance among the
sampled cities.

Page 8 of 7