Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 11

EXERCISE 2

Submitted to- Dr. inderjeet oberoi

Submitted By- Akshay Kushwaha


401508007
Q 1.

Future price seems to be increasing and it’s positively skewed because mean is greater than
median. We can also see a kind of structural break in the end around 11 on the X axis. It started
increasing in the middle of 10 and 11. And after 11 it changed its level and shifted to a higher
level.

Spot price also seems to be increasing and positively skewed because mean is greater than
median. We can also see a kind of structural break in the end around 11 on the X axis. It started
increasing in the middle of 10 and 11. And after 11 it changed its level and shifted to a higher
level.
Descriptive Statistic
Future Spot
Mean 2072.001 2037.214
Median 1846.500 1813.425
Maximum 5255.000 5098.450
Minimum 1024.000 1005.700
Std. Dev. 713.9555 711.7637
Skewness 2.024124 2.024238
Kurtosis 7.248589 7.230653

Jarque-Bera 3125.322 3111.689


Probability 0.000000 0.000000

Sum 4512818 4437053


Sum Sq. Dev. 1.11E+09 1.10E+09

Observations 2178 2178

The null hypothesis for the Jarque-Bera test is that the data is normally distributed; the alternate
hypothesis is that the data does not come from a normal distribution. Thus, this table of
descriptive statistic majorly tell us about the normality of the given data.

Interpretation

1. Future price data does not come from a normal distributed sample because the
probability value is less than 0.05 i.e. I can reject the null hypothesis.
2. Spot price data does not come from a normal distributed sample because the probability
value is less than 0.05 i.e. I can reject the null hypothesis.
Q2

The t-value measure the size of the difference relative to the variation in your sample data. In
another word, T is simply the calculated difference represented in units of standard error. The
greater the magnitude of T, the greater the evidence against null hypothesis.

Model interpretation

1. Constant C has p-value for test less than 0.05 therefore we can say that this variable is
significant and the true mean is not equal to comparison mean.
2. And for the variable spot also the p-value is less than 0.05 therefore we can say that this
variable is significant and the true mean is not equal to comparison mean.
Q3
1) Normality Test:

The null hypothesis for the Jarque-Bera test is that the data is normally distributed; the alternate
hypothesis is that the data does not come from a normal distribution. Thus, this table and graph
of Normality test majorly tell us about the normality of the given data.

Interpretation

1. The p-value is less than 0.05 therefore we can reject null hypothesis.
2. The data does not come from a normality distribution sample.

2) Zero Mean of the Disturbance:


Here as we can see in above graph, mean of residual is -2.16e-13 i.e. near to
zero. So it is fulfilling the OLS zero mean of disturbance assumption.
3) Serial Correlation LM test:

The Breusch–Godfrey test is a test for autocorrelation in the errors in a regression model. The


null hypothesis is that there is no serial correlation of any order and alternate Hypothesis is that
there is serial correlation.

Interpretation

1. The p-value of chi-square is less than 0.05 therefore we can reject the null hypothesis.
2. There is serial correlation.
4) Heteroskedasticity Test:

The Breusch-Pagan-Godfrey Test is a test for heteroscedasticity of errors in regression model.


The null Hypothesis is that there is constant variance and alternate Hypothesis is that there is no
constant variance.

Interpretation

1. The p-value is less than 0.05 therefore we can reject the null hypothesis.
2. There is no constant variance.

5) No correlation between regressor and disturbance:


Null hypothesis: True auto correlation is lesser than or equal to 0
With the correlation between X (spot) and resid is 6.15e^15 which is 0.0000188189 which is
very close to zero so we can say both are not correlated. This assumption is fulfilled.

6) Value of Xi is a known constant in the probability distribution of Y:

If the means of error term is zero we can assume that spot price is not random. So this
assumption holds true. 

Q 4.
Re-estimating the model

As Unit root of both the series exist, we created the difference series for both of the variable i.e.
dfuture and dspot to overcome the unit root problem. With the difference series of spot and
future unit root is improved and we see that now p-value is less than 0.05 so we can say that the
stationary of model is also improving. So now we will proceed with the estimating equation.
Model B : dfuture c dspot

Here we can see the the F-stats value is 0.00000 which is statistically significant. But we need to
check other OLS assumptions too.
Residual diagnostic output 
1) Normality:

2) Breusch-Godfrey Serial Correlation LM Test:


3) Heteroskedasticity Test: Breusch-Pagan-Godfrey:

Q5

Interpret all the results:

Model equation future c spot is directly rejected because of unit root. Now we can check the
model with the differenced series dfuture c dspot.

With this model our output for the assumptions test also changes. Now correlation exist and out
covariance in error terms is also same but the series are still not normally distributed. So we
conclude that out model is not a good fit.

You might also like