[Type the document title]

[Type the document subtitle]

[Type the abstract of the document here. The abstract is typically a short summary
of the contents of the document. Type the abstract of the document here. The
abstract is typically a short summary of the contents of the document.]

Haier

[Year] [Type the company name]

[Pick the date]
Question 1:

Estimation Command:

=========================

LS NETTFA C AGE INC FSIZE MARR MALE E401K

Estimated Equation:

=========================

NETTFA = C(1) + C(2)*AGE + C(3)*INC + C(4)*FSIZE + C(5)*MARR + C(6)*MALE +

C(7)*E401K

NETTFA = -54.695356845 + 1.02415124639*AGE + 0.982000518598*INC -

1.6790359193*FSIZE - 6.67224005606*MARR - 0.384393227632*MALE +
5.77179206399*E401K
Part a:

The model involved 6 variables involving coefficients while one intercept. The coefficients of
age, income and dummy variable for retirement pension plan have positive coefficients while
remaining 3 have negative coefficients. positive coefficients of age, income and pension
variables are indicating the presence of direct relationship among these variables. As, it seems
the increase in age, income and pension plan would cause an increase in net financial assets.
While among these three variables, pension plan has larger effect than other two. However,
negative impact of family size, married and male respondent is visible due to negative
coefficients of all of them. It can be interpreted as increasing family size having more married
male members have less financial assets.

Significance of coefficients:
The p values for all of the coefficients are small enough that seems all of them are playing their
role significantly in prediction of financial assets except males. As, p value for male is larger (i.e
0.8145 > 0.05). so, it indicates males are not playing their role significantly in predicting
financial assets.

Intercept:

The intercept of model is -54.69 which can be interpreted as if all of the variables are removed
than also financial assets would be decreased by a factor of 54.69 units.

Part b:

Null and Alternative Hypothesis:

Ho: There is no significant impact of age on financial assets.

Ha: There is significant impact of age on financial assets.

Mathematically;

Ho: β 1=0

Ha: β 1 ≠ 0

Level of significance:

Alpha = 0.05

Test Statistics:

^β−β
t=
s .e ( β^ )

Critical Region:

We would reject null hypothesis if

P value < alpha

Calculations:
The p value calculated for coefficient of Age is 0.0000 which is very small (i.e < 1%, 5% and
10%).

Conclusion:

As, p value for coefficient of Age is smaller than all significant levels. So, it is providing
significant evidence for rejection of null hypothesis concluding the variable Age is significantly
impacting in financial assets.

Question 2: Joint Significance test:

Part a:

Null and Alternative Hypothesis:

Ho: All of the independent variables are not jointly significantly impacting on financial assets.
Ha: All of the independent variables are jointly impacting on financial assets.

Mathematically;

Ho: β i=0

Ha: β i ≠ 0

While i=1,2,3,4,5,6

As, there are 6 independent variables in the model.

Level of significance:

Alpha = 0.05

Test Statistics:

Critical Region:

We would reject null hypothesis if

P value < alpha

|F cal|> F tab

|F cal|> F α (v , v )1 2
2

|F cal|> F α (k−1 ,n−k)

2

|F cal|> F 0.05 (7−1 ,9275−7)

2

|F cal|> F 0.025(6 ,9268)

|F cal|>1.72
Calculations:

The p value calculated for F test is 0.0000 which is very small (i.e < 1%, 5% and 10%).
However, calculated value of F statistics is 331.67.

Conclusion:

As, p value for F test is smaller than all significant levels with larger F test value of 331.67(i.e
>1.72). So, they are providing significant evidence for rejection of null hypothesis concluding
that all of the independent variables are jointly impacting on financial assets.

Part b:

The F test is conducted for exploring overall significance of model by means of joint impact of
all of the independent variables on dependent variable. Here, f test is providing enough evidence
of significance of model. It is exploring that jointly all of the variables are impacting
significantly. However, t test determines individual significant impact of all variables. One of the
variable is not impacting significantly on financial assets. So, it seems that if we take separately
than that particular variable may not cause significant impact on financial assets while as a
combined impact the variation is significant. this may be due to the reason that there exists
multicollinearity, irrelevant variable or small size.

Question 3: Testing restrictions:

Part a:

Restriction:

1
2∗age= ∗fsize
2

Multiplying both sides by 2:

4∗age=fsize

4∗age−fsize =0

After running the regression model. The restriction would be implemented.

Part b:

Wald test Manually:

( 4∗1.024−−1.679 )2
W=
42 var ( ^
β age ) + var ( ^
β fsize )

( 4∗1.024+1.679 )2
W=
16∗0.0595 2+ 0.48922
 33.3506
W=
0.2959

W =112.71

Part c:

The value of wald statistics is very large with smallest p value providing enough evidence of
possible rejection of null hypothesis concluding significant evidence that two extra years of age
has half the effect of one more family member, on nettfa.

Question 4: Multicollinearity test:

Part a:

Multicollinearity is the technique for exploring presence of relationship among independent

variables. It can be observed by means of variance inflation factor (i.e VIF). If VIF of two
variables is more than 5 than there is presence of high multicollinearity among variables while
for less than 5 there is no problem of multicollinearity. Here, VIF for both of these variables is
small that seems no significant evidence is present for predictions of multicollinearity.

Part b:

Part i:

At the point when multicollinearity exists among factors, it can swell standard mistakes,
debilitate the measurable force of tests, and produce questionable coefficient gauges. Settling this
issue requires procedures that rely upon the specific situation and objectives of the examination.

Determining Multicollinearity:

 VIF
 Correlation matrix
 Eigen value index

After exploring multicollinearity it can be removed using:

 Removal of multicollinear variables.

 Join collinear variables
 Standardizing collinear variables.
 Principal component analysis
 Regularization technique

If multicollinearity is present than after removal of multicollinear variable the goodness of fit and
prediction of dependent variable improves.

Part ii:

Multicollinearity influences the exhibition of the Customary Least Squares (OLS) assessor in
relapse examination, yet it doesn't abuse the suspicions of the OLS model itself. Here are the
critical results of multicollinearity for the OLS assessor:

1. It increase the variance of estimates of coefficients of regression model. In presence of

them the estimated model is not accurate.
 2. The coefficients are tested by means of t test which would be reduced to zero having
large standard error. Thus, no coefficient of independent variable impacts significantly on
dependent variable.
3. The impact of most significant variable among all independent variables would not be
estimated.
4. Precision of model is reduced.
5. The estimated model can’t be considered as best linear unbiased estimator.

Question 5: Heteroskedasticity: Graphical Analysis:

Part A:

The scatter plot indicates no significant pattern has been observed among values. So, no variation
exists among values. Thus, there is no heteroskedasticity among residuals and fitted values.

Part b: Consequences of heteroskedasticity:

Heteroskedasticity alludes to a circumstance where the change of the mistakes (residuals) in a
relapse model isn't steady across all levels of the free variable(s). While it doesn't predisposition
the coefficients assessed by Customary Least Squares (OLS), it influences the proficiency and
unwavering quality of the model in more ways than one. The following are the vital outcomes of
heteroskedasticity on the OLS assessor:

 The results appeared unbiased but there are not efficient estimator of OLS model.
 The estimated standard errors either under or over estimate the results.
 A true hypothesis may sometimes be rejected due to inaccurate standard errors.
 The confidence interval of coefficients become too narrow or too wide.
 Prediction and forecasting can be more effective.

Question 6: White’s test:

Auxiliary Regression:

The regression that is estimated after observing multicollinearity, heteroskedasticity and

autocorrelation. Here, auxiliary regression is estimated by taking squared residuals as dependent
variable while all other variables similar to original regression were considered as independent
variables.

Null and Alternate Hypothesis:

Ho: No Heteroskedasticity is present among variables leading to constant variance of squared

residuals and all of the independent variables.
Ha: Heteroskedasticity is present among variables leading to constant variance of squared
residuals and all of the independent variables.

Test Statistics:

2
W =n . R

Calculations:
Conclusion:

The p value for white test is small providing significant evidence of rejection of null hypothesis
concluding the Heteroskedasticity is present among variables leading to constant variance of
squared residuals and all of the independent variables.

Part b:

Preference of white test on Breusch pagan Godfrey test:

The White’s test is preferred over Breusch pagan Godfrey test due to the following reasons:

1. Flexibility
2. Robustness
3. Sensitivity

Flexibility:

The White’s test is considered more flexible than Breusch pagan test as it can be applied to
regression of any type either linear, non-linear or complex but Breusch pagan can only be
applied to linear models.

Robustness:
The White test can be applied whether data is heteroskedastic while in that case Breush pagan
can’t be applied.

Sensitivity:

White’s test can be applied when interaction term is involved between model rather Breush
Pagan test.

Question 7:

At the point when heteroskedasticity is available in a relapse model, it implies that the
fluctuation of the blunder term isn't consistent across all perceptions. If heteroskedasticity is of
the following form:

1
σ 2ui =σ 2u . fsiz e 2

Thus, there is direct relationship among variance and family size. As, family size changes so it
means that variance of error is not constant which is termed as presence of heteroskeasticity.
Thus, firstly removal of heteroskedasticity is required.

The heteroskedasticity can be removed by applying transformation using standardized technique.

Transformation can be applied in following way:

' y
y=
√ fsize
' x
x=
√ fsize
So the transformed model would be:

' '
y =β o + β 1 x +e i

Question 8:

Part a:
The model estimated using White’s test explored impact of all of the coefficient increases while
p value for two of them are indicating non-significant evidence. Thus, using white’s test males
and family size are not impacting significantly on financial assets.

Part b:

In the following cases White standard error can be applied:

 If someone is suspect about presence of Heteroskedasticity.

 Form of Heteroskedasticity is not known.
 Determination of best statistical inference technique.
 Large no. of independent variables are included in the model.
Question 9: Autocorrelation:

Part a:

Here the plot indicates no linear pattern has been observed among 1 year lagged series. In case of
no linear pattern there is no presence of autocorrelation.
Question 10: Durbin Watson test:

Durbin Watson test is statistical technique applied for determination of autocorrelation among
residuals of fitted model. Thus, for determining one order autocorrelation among residuals and
independent variables Durbin Watson test is best choice.

Null and Alternate Hypothesis:

Ho: There is no autocorrelation present among residuals and independent variables of model.

Ha: There is autocorrelation present among residuals and independent variables of model.

Here value of Durbin Watson test is 1.93 which is less than 2 so it seems small positive
autocorrelation is present. Athough it is nearly equal to 2 after rounding off than there would be
no autocorrelation.
Question 11: Functional form:

Remesy test:

The best fitted model can be observed by fitting various models and than comparison of their
dagnostic tests. The model having best diagnostic techniques would be considered as best. So,
three different types of model involving semi logarithmic, quadratic and model having
interaction terms.
Semi logarithmic model:
Quadratic model:
Model involving interaction term:
After comparison of R square of all models, the simple linear regression is best one.

Question 12:

Semi logarithmic model:

Question 13:

Estimation Command:

=========================

LS RESID_SQ C AGE INC FSIZE MARR MALE E401K IS

Estimation Equation:

=========================

RESID_SQ = C(1) + C(2)*AGE + C(3)*INC + C(4)*FSIZE + C(5)*MARR + C(6)*MALE +

C(7)*E401K

Substituted Coefficients:

=========================
RESID_SQ = -12104.4649751 + 171.587511252*AGE + 299.728159583*INC -
92.1852250809*FSIZE - 3656.65850913*MARR - 242.734870668*MALE -
1857.22325139*E401K

𝐻0: 𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙𝑠 𝑎𝑟𝑒 𝑛𝑜𝑟𝑚𝑎𝑙𝑙𝑦 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑒𝑑

𝐻1: 𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙𝑠 𝑎𝑟𝑒 𝑛𝑜𝑡 𝑛𝑜𝑟𝑚𝑎𝑙𝑙𝑦 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑒𝑑

Test for statistics and distribution:

[ ]
2
s2 ( K −3 )
JB=n + X 21−α , 2
6 24

The p value for jarque bera test is small enough providing evidence that residuals are non-
normally distributed.

However, distribution is positively skewed as observed from histogram as well.

Question 14:
The analysis involved distribution as well as prediction of financial assets by means of various
other variables. It contained data of 9275 individuals from the Survey of Income and Program
Participation (SIPP) of 1991. The results explored that financial assets are depending upon age,
income and pension plan positively while on gender, family size and marital status positively.
So, it means that financial assets of single females having less family size increases. Thus, it will
help the company to explore its previous assets and predictions of assets in future as well. So, as
to apply those techniques that may improve economic condition. Further, it will help to remove
those factors that are lowering the financial assets. Also, the results explored the positive
distribution has been observed that seems by increasing time the assets are increasing. In order to
improve their impact positively some other measures such as family control program and
increase in income would be more helpful.