Beruflich Dokumente
Kultur Dokumente
Roll # :
17010037
Semester:
Course Title:
Academic
Year
2015 - 2016
Instructor:
Dr Farrah Arif
Date
Exam:
Final
Time Allowed:
4 hours
Total Marks:
100
The instructions below must be followed strictly. Failure to do so can result in serious grade
loss.
Questions
Q1: Based on linear trend equation, what is the estimated price of an apartment in 2010? (Hint: write
down the model; interpret the regression equation).
Using linear trend equation, find the growth rate of the apartments prices. Do you need any
transformation to find the growth rate?
Does this model violate OLS assumption(s)?
Table 1
Table 2
From table 2,
Ln (Price) = 8.055+ .002(size) -.015 (Age) + 0.337 (Green) + 0.060 (MGT Fee) -0.314 (Y97) -0.243(Y98) -0.118 (Y99)
-0.211 (Y00) 0.071(Y01) -.073(Y02) + .490 (Y04) +.589(Y05) +0.578 (Y06) +0.774 (Y07) + .927(Y08)
Price of apartment will increase by 0.2% with change in size of department when all other factors are held constant mean
for same green ratio, age and year of transaction, but it will have major effect of 33% increase in price when increase one
unit of green area in residence. Price also decrease by increase in age of apartment. As age of same type and size of
apartment will increase by one year, price will be decrease by 1.5%. From table 2 it is also clear that some of the year
variables are not significant and they will have negligible effect on our model.
From above linear trend equation it is clear that price of a same type of apartment will increase by 105 110 % with
reference to year 2003.
From above equation it is clear that price of same size and type of apartment is 31% lower in 1997 with reference to year
2003 and increase by 92% in year 2008 when age, size and green ration is same in different years.
OLS Assumptions:
Linear Relation:
From above figures it is clear that size of apartment has direct relation with price, but there is no
relation between price and green area but we will include this in our model because it make economic
sense.
Normal Distribution:
Multi collinearity
From table 2, it is clear that all VIF values are less than 10 so there is no multi collinearity.
Auto Correlation:
From table 1, D Watson value is 1.29 which lies in auto correlation region which is also clear from
below graph.
Figure 1
Hetroscadicity:
Figure 2
From above figure 2, it is clear that there is no hetroscadicity with in this model.
Q2: At .05 level of significance, is there a difference in the mean price of an apartment whether it is
close to bus station? At .05 level of significance, is there a difference in the mean price of an
apartment whether it is close to main road? What inference can be drawn from the findings?
Table 2 - 1
Table 2 - 2
Form table 2-2 it is clear that those for a same type, size and age apartment which are near to bus station will
have 36 % higher prices than those apartments which are not close to bus station.
2-B
Table 2 - 3
Table 2 - 4
From table 2-4 it is clear that same type of apartment which are close to road are 58.4% costly than those
which are not close to main road.
Q3: Develop a correlation matrix of the variables that Mr. Dhelvi identified. Which variables would
you select for your analysis? Why? Why not?
Table 3- 1
Price of an apartment directly depends upon size of property so this is a very important factor
we cannot neglect this factor.
Age of a property also plays an important role as property will get older is will be more out of fashion
and more far from latest design so value of property will reduce with age.
Monthly management fee is directly link with better services of area so it also has an important
impact on price of property.
According to new fashion trend new buildings have more green area in house and compulsory by
some of residential colonies so green area also has impact on price.
Year of transaction is basically gives economic condition, value of property and money in that period
so also very important factor.
Q4: Determine the best regression equation, using Price as the dependent variable and the other
(minus years) as independent variables. Would it be a useful model? Why? Why not?
Table4- 1
Table4- 2
Regression equation:
Price = -2608.33 + 46.388 (Size) + 304.447 (Age) + 777.019 (MGT Fee) + 5726 .981 (Green) + 2041 .073 (Main Road)
212.208 (Bus) + 2152.730 (Lightrail)
From table 4-1 it is clear that different models were tested, with different combinations with adding each
individual dummy variable Adjusted R2 improved. Final model explain 83.5 % variation in price variation
with these variables.
This model is not a quite useful model although it explains 83.5% variation in price but some factors does not
make any economic sense like as apartment gets older it value will decrease but according to this model price
will increase with increase in age of property.
Also those houses which are near to bus station will have lesser prices than those which are far from bus
station this also does not make any sense, more over it makes variable BUS in significance but this variable is
more important for us as we have seen in question 2 that for similar type of house 36% prices difference for
those house which are near to bus stop so we cannot neglect these factors.
Q5: Do you need any transformation(s) to improve the model? If yes, explain.
As mention above, some sign does not make any economic sense so we need to transform some data.
Table 5- 1
Figure 5- 1
From table 5-1, it clear that with transforming data from price to Ln (price) improves
value of Adjusted R2 from 0.835 to 0.947 also when we check relation between Ln
(price) and size it seems that it best explain by quadratic equation, it changes R2 from
0.571 to 0.714 as mention in figure 5-1. Also from table 5-1 it shows that addition of Sqr
of size give a jump of Adj R2 from 0.947 to 0.957.
Q6: Determine a possible best model (given the limited data) that can be used for the webportal. (Hint: give explicit details including assumptions of OLS.)
Table 6- 1
Model Summaryb
Model
R Square
.920a
Adjusted R
Square
Estimate
.846
.841
Durbin-Watson
.21685
.790
Coefficientsa
Standardized
Unstandardized Coefficients
Model
1
B
(Constant)
Coefficients
Std. Error
Beta
8.170
.107
Size
.003
.000
Age
-.120
Mainroad
Collinearity Statistics
t
Sig.
Tolerance
VIF
76.305
.000
.285
6.935
.000
.428
2.339
.037
-.378
-3.267
.001
.054
18.504
.415
.053
.360
7.896
.000
.349
2.865
Bus
.127
.043
.106
2.962
.003
.562
1.781
MGTFee
.110
.016
.314
6.680
.000
.327
3.056
Lightrail
.172
.039
.159
4.371
.000
.549
1.820
Age2
.016
.004
.531
4.615
.000
.055
18.261
Table 6- 2
Ln (Price) = 8.160 +0.003(Size) 0.120 (Age) + 0.415 (Main Road) + 0.127 (Bus) +0.110 (MGT Fee) +
0.172 (light Rail) + 0.016 (Age2)
From table 6-1 it is clear that, our Adj R2 slightly decline from our base model but it has make our some of
major factors significant like Bus which we have already seen individually has impact of 36% change in
price.
As we have earlier mention that value of property decline with increase in age of property, from above
equation it is clear that value of property is increasing at decreasing rate for 1 st 3.75 year it will increase but
after that it will start decreasing, this thing make an economic sense because property is new and according to
upcoming latest styles but as new trends arrived value will start declining.
OLS Assumptions:
Normal Distribution:
From below figure it is clear that residuals folooe normal distribution.
Figure 6- 1
Linear Relation:
There is a linear relation between dependent and independent variable
From table 6-2 it is clear that VIF value is less than 10 for all variable except Age and age2 which are directly
co related with each other.
Auto Correlation:
From table 6-1, D Watson value we found that 0.79 which lies in auto correlation region so we can say that
residuals are correlated with each other.
Hetroscadicity:
Figure 6- 2
ANOVAa
Model
1
Sum of Squares
Regression
Residual
Total
df
Mean Square
54.916
7.845
9.969
212
.047
64.885
219
Sig.
166.829
.000b
Q7: Over the years, prices of the apartments have increased. Make a graph, showing the increase in
price per year? (Hint: Using a regression model). Explain.
Coefficientsa
Standardized
Unstandardized Coefficients
Model
1
B
(Constant)
Coefficients
Std. Error
Beta
-2445520.350
159957.502
1224.608
79.760
Year
.721
-15.289
.000
15.354
.000
Model Summary
Model
1
R Square
a
.721
.520
Adjusted R
Square
Estimate
.517
Sig.
3582.01244
Figure 7- 1
From figure 7-1, it is very much clear that price of apartment is increasing by years, for every year
increasing in price indicates that for every year price is different for different age and different size.
f(x) = 1224.61x + 0
12000
10000
8000
6000
4000
2000
0
10
12
14
Also not restrict himself for selling and buying, but also work on commercial areas rental business, as
economy is improving so it will also have major impact on his business improvement.
Year
2010
2010
2010
Quart
er
I
II
III
2010
IV
6.25
6.125
2011
6.75
6.5
2011
II
12
7.25
2011
III
7.5
7.375
2011
IV
7.75
7.625
2012
8.75
8.25
2012
II
16
9.5
9.125
2012
2012
III
IV
12
4
9.5
9.5
Absents
4
10
7
2010
I
II
III
IV
0.4897
96
Movin
g Avg
CMA
Ab/CM
A
6
0.4897
96
0.7692
31
1.7142
86
1.2203
39
0.5245
9
0.7272
73
1.7534
25
1.2631
58
2011
2012 Avg
SI
0.7692 0.7272
0.7073
31
73
0.75
91
1.7142 1.7534
1.6391
86
25
1.73
73
1.2203 1.2631
1.1739
39
58
1.24
39
0.5245
0.4794
9
0.51
96
Total
4.23
4
Correction
Factor
0.95
SI
0.707
1.639
1.174
0.479
0.707
1.639
1.174
0.479
0.707
1.639
1.174
0.479