You are on page 1of 4

COMMON MISTAKES MADE IN TUTORIAL 4

Question 1 Correlation Analysis



1) The Pearson Correlation value obtained is 0.664. It indicates that there is a weak upward slop
of regression line, having a weak positive linear relationship between Price and Year as the
value is not close to absolute value of 1. THIS IS MODERATELY STRONG.

2) This indicates a large positive value which is a clear upward slope of the regression line. This
concludes that Singapores landed property prices will increase as the years go by. ONLY
INTERPRETED THE SIGN (+/-); MISSING STATEMENT OF THE STRENGTH OF
RELATIONSHIP WHICH IS THE ESSENCE.

3) At 1% significance level, a large negative value (-0.8) indicates a clear downward slope of
the regression line between the variables of type and area. This indicates that the
correlation between the variables of type and area is weak.

A NEGATIVE SIGN FOR CORRELATION DOES NOT MEAN THE RELATIONSHIP IS
WEAK.

4) Need to interpret the Sig. (or p-value) of the Correlation coefficient. - MISSING NULL &
ALTERNATIVE HYPOTHESIS.

5) Ho: There is no correlation between price and year
H1: There is correlation between price and year
NEED TO STATE NULL AND ALTERNATIVE HYPOTHESIS IN A MATHEMATICAL FORM.
BUT AS A CONCLUSION OF YOUR STATISTICAL TEST, YOU CAN EXPLAIN IT IN
WORDS.

6) The population correlation coefficient is hypothesized to be equal to 0. The null hypothesis states
that the population correlation coefficient does not equate to 0. Therefore, with a significance level
of 1%, we can reject the null hypothesis. A CARELESS MISTAKE!


Question 2 Stepwise Regression

7) For the results generated by Stepwise procedure, the last set of model results belongs to the
best-fit model. You only need to interpret those and ignore the rest. LOOK AT THE LAST SET
OF RESULTS FOR THE BEST-FIT MODEL.

8) Need to state the hypothesis for testing the significance of F-ratio, t-test. MISSING NULL &
ALTERNATIVE HYPOTHESIS.
9) The hypothesis is as follows:
H
0
: = 0
H
1
: is not equal to 0

WRONG HYPOTHESIS FOR TESTING F-RATIO!

10) H0: p = 0
H1: p 0

WRONG HYPOTHESIS FOR TESTING t-statistic!

11) The F-ratio test is based on:
H0: There is no significant relationship between dependent variable Y (Price) and the independent
variables X (Type, Year and Area)
H1: There is significant relationship between dependent variable Y (Price) and the independent
variables X (Type, Year and Area)
WRONG STATEMENT OF THE NULL & ALTERNATIVE HYPOTHESIS! IT SHOULD BE ABOUT THE
VARIABLES STATISTICAL CONTRIBUTION TO EXPLAINING Y (OR THE MODEL).
12) F ratio

Independent X variable, type of property, should be removed from the test given that it is highly
correlated to the gross floor area and would distort the regression values. The decision rule is if the
calculated statistical value F*> than 3.04 of the F distribution with 2-1 and 204-2 degrees of freedom,
we reject H0 that states that there is no correlation between y variable, price of property and th x
variables, gross floor area and the year of sale.

THE F RATIO TEST IS NOT ABOUT CORRELATION.

13) Since F-ratio =193.222 > F
200,3,(203-3-1)
=2.696, we reject and conclude that at least one of the
independent variable is related to the dependent variable Price.

YOU NEED TO BE CLEAR WHAT ARE YOU REJECTING? NULL?





14) The t-test is based on:
H0: There is no significant relationship between dependent variable Y (Price) and the independent
variables X (Type, Year and Area)
H1: There is significant relationship between dependent variable Y (Price) and the independent
variables X (Type, Year and Area)
WRONG STATEMENT OF THE NULL & ALTERNATIVE HYPOTHESIS! IT SHOULD BE ABOUT THE
VARIABLES STATISTICAL CONTRIBUTION TO EXPLAINING Y (OR THE MODEL).
15) Price = -2848227.807+ ( 42109.990( year) --131501.678+268.199 )+ ei

WRONG EQUATION!

16) Y = 42109.990 X
1 +
(-131501.678 X
2
) + 268.199 X
3
+ (-2848227.807) + e, where e is the random error
in Y

NEED TO STATE WHAT IS X
1
, X
2
AND X
3
.
17) Price = -2848227.8(constant) + 42109.990(year) + 268.199(GFA) + e

Independent x variable, type of property should be removed from the test given that is highly
correlated to the GFA and thus, would distort the regression values.

THERE IS NO NEED TO CORRECT FOR MULTICOLLINEARITY AFTER YOU RUN REGRESSION !

18) Price = 268.199(area) + -131501.678(type) + 42109.990(year) + constant
NEED TO STATE THE VALUE OF CONSTANT AS WELL.
19) Price = -2848227.807 + 42109.990 131501.678 + 268.199 + e

INCOMPLETE EQUATION - NEED TO INCLUDE THE X
1
, X
2
, X
3
.

20) H
0
:
1
=
2
=
3
= 0
H
1
:
i
= 0 (at least one i)

H
1
SHOULD HAVE BEEN POPULATION PARAMETER NOT EQUAL TO ZERO.

21) Demand = -2848227.807 (constant) + 42109.990 (year) -131501.678 (type) +268.199 (area) +e

THE QUESTION IS ABOUT PREDICTING PROPERTY PRICE NOT DEMAND!



22) Model 1. Price = -2741549.184 + 38701.644 (Year) + e
Model 2. Price = -2484715.669 + 41068.369 (Year) - 205509.487 (Type) + e
Model 3. Price= -2848227.807 + 42109.990 (Year) - 131501.678 (Type) + 268.199 (Area) + e
THERE ARE NO MODEL 1, 2 & 3. THE STEPWISE METHOD PRODUCES THE BEST-FIT MODEL IN THE
LAST STEP.

You might also like