You are on page 1of 6

Michael Sandy Moore Homework #7 DADSS

1. My best friend here at school has a problem: he can’t find the right girl. After two years being
one of the biggest players that CMU has ever seen, he figured it was time to try settling down a
bit and start dating. The trouble is, he can’t figure out whom to date. In the past year, John has
grown closest with three girls: Molly, Niamh (pronounced “neeve” – it’s Irish), and a girl he met
on spring break that goes to University of Connecticut – for simplicity’s sake I’ll just call her
UConn girl. Molly, Niamh, and UConn girl can be judged on three different attributes for a
Multi-attribute decision analysis: Personality, attractiveness, and income.

First, John needs to make sure that dating all of these girls does not violate any assumptions. He
agrees to obey the law of transitivity (i.e. if Molly is preferred to Niamh, and Niamh is preferred
to UConn girl, then Molly is preferred to UConn girl). He agrees to obey the idea of preferential
independence (i.e. the attractiveness of any of the girls is not influenced by their personality or
attractiveness). He agrees to obey the idea of difference independence (i.e. if he prefers a girl
with a great personality to a girl with a good personality if she is attractive, he must maintain
this preference even if she is very attractive). He agrees to obey the idea of trade-off
independence (i.e. if he would trade an attractive girl for one with a great personality, the
decision to do so cannot be influenced how much income the girl makes).

Second, John now needs to rank each alternative for the girls’ attributes in importance to him.
He does it like so:

Attractiveness: V(Very attractive) = 100. V(Attractive) = 50. V(So-so) = 0.


Personality: V(Great) = 100. V(Good) = 80. V(So-so) = 0.
Income ($/wk): V($50) = 100. V($45) = 90. V($40) = 70. V($35) = 30. V($30) = 10. V($25) = 0.

Thus, the value functions for each would look like so:
Michael Sandy Moore Homework #7 DADSS

Third, John must assess the scales in which he wants to make his decision to date. There are
two methods he can use: the swing weight method and the direct-tradeoff method.

Swing Weight Method


John isn’t a completely shallow person, so he ranks these attributes in importance like so:
Personality = 100
Attractiveness = 60
Income = 20

To normalize this, we divide each of these swing weights by the total value that the weights add
up to. Thus:

Personality = 100 / (100 + 60 + 20) = 0.5556


Attractiveness = 60 / (100 + 60 + 20) = 0.3333
Income = 20 / (100 + 60 + 20) = 0.1111

Direct-Tradeoff Method
Another way for John to assess his decision is through direct-tradeoff, which involves asking
John what we would levels of our continuous variable (income) he would trade for
improvements or detractions from our categorical variables (attractiveness and personality).
Thus:

Attractiveness (we leave out personality and assume it’s constant)


Option A1 = ($50, So-so) and Option A2 = (?, Very Attractive). We ask John how much lower of
an income is he willing to deal with to make sure that his future wubby is very attractive. John
says he would have no preference if she was making $30/wk.

Personality (we leave out attractiveness and assume it’s constant)


Option A3 = ($50, So-so) and Option A4 = (?, Great). We ask John much lower of an income is he
willing to deal with to make sure that his future wubby has a great personality. John says he
would have no preference if she was making $10/wk.

We use these two discoveries to set up this equation for V(A1) = V(A2) when the income of A2 is
$30/wk:

WxVx($50) + WyVy(So-so) = WxVx($30) + WyVy(Very Attractive)


Using the value functions on the previous page, we can simplify this equation to:
100Wx + 0Wy = 10Wx + 100Wy
Thus, Wy = 0.9Wx
Michael Sandy Moore Homework #7 DADSS

We can also set up this equation for V(A3) = V(A4) when the income of A4 is $10/wk:

WxVx($50) + WzVz(So-so) = WxVx($10) + WzVz(Great)


Using the value functions on the previous page, we can simplify the equation to:
100Wx + 0Wz = 0Wx + 100Wz
Thus, Wz = 1Wx

To normalize this, we must set Wx + Wy + Wz = 1


Then, we can substitute the earlier equations to get:
Wx + 0.9Wx + 1Wx = 1.0, or Wx = 0.345

We can then substitute this new value back into the old equations to find their tradeoff weights:
Wy = 0.9 x 0.345 = 0.311
Wz = 1 x 0.345 = 0.345

To sum up both the swing weights and the tradeoff weights, we can use this grid:

Attribute Worst Best Rank Swing Weight Normalized Weight (Swing) Tradeoff Weight

X = Income $25 $50 3 20 0.1111 0.345


Y = Attractiveness So-So Very 2 60 0.3333 0.311
Z = Personality So-so Great 1 100 0.5556 0.345

Additionally, the breakdown of the girls looks like this:

Attributes Molly Niamh Uconn Girl

Income $35 $30 $50


Attractiveness Attractive Very Attractive So-so
Personality Great Good So-so

Thus we can compute the values for each prospective girlfriend using swing weights first:

V(Molly) = (0.1111 x 30) + (0.3333 x 50) + (0.5556 x 100) = 75.6


V(Niamh) = (0.1111 x 10) + (0.3333 x 100) + (0.5556 x 80) = 78.9
V(UConn) = (0.1111 x 100) + (0.3333 x 0) + (0.5556 x 0) = 11.1

In the case of these swing weights, it appears that Niamh is the lucky girl by a narrow margin.
Michael Sandy Moore Homework #7 DADSS

Furthermore we can compute the values for each prospective girlfriend using Tradeoff weights:

V(Molly) = (0.345 x 30) + (0.311 x 50) + (0.345 x 100) = 60.4


V(Niamh) = (0.345 x 10) + (0.311 x 100) + (0.345 x 80) = 62.2
V(UConn) = (0.345 x 100) + (0.311 x 0) + (0.345 x 0) = 34.5

Again, in the case of Tradeoff weights Niamh is the lucky girl by an even narrower margin.

Finally, we can conduct a sensitivity analysis that compares the girls when John’s preferences
are changed and when we weight all attributes equally. Setting all the weights equally would
yield this:

V(Molly ) = (0.3333 x 30) + (0.3333 x 50) + (0.3333 x 100) = 60


V(Niamh) = (0.3333 x 10) + (0.3333 x 100) + (0.3333 x 80) = 63.3
V(UConn) = (0.3333 x 100) + (0.3333 x 0) + (0.3333 x 0) = 33.3

Again, Niamh wins by a narrow margin.

When you change John’s attractiveness preferences, it looks like this:

John values attractiveness at 40 (2/5 as much as personality):


V(Molly) = (0.125 x 30) + (0.25 x 50) + (0.625 x 100) = 78.8
V(Niamh) = (0.125 x 10) + (0.25 x 100) + (0.625 x 80) = 76.3
V(UConn)= (0.125 x 100) + (0.25 x 0) + (0.625 x 0) = 12.5

John values attractiveness at 80 (4/5 as much as personality):


V(Molly) = (0.1 x 30) + (0.4 x 50) + (0.5 x 100) = 73
V(Niamh) =( 0.1 x 10) + (0.4 x 100) + (0.5 x 80) = 81
V(UConn) = (0.1 x 100) + (0.4 x 0) + (0.5 x 0) = 10

In both cases, Niamh still gets the privilege of dating John.


Michael Sandy Moore Homework #7 DADSS

This table can sum up the sensitivity analysis and the entire dating issue for John:

Molly Niamh Uconn Girl Rel. Weight 1 (orig.) Swing Weight 1 (orig.)

Income $35 $30 $50 20 0.1111


Attractiveness Attractive Very Attractive So-so 60 0.3333
Personality Great Good So-so 100 0.5556

Rel. Weight 2 Swing Weight 2


Income 20 0.125
Attractiveness 40 0.25
Personality 100 0.625

Rel. Weight 2 Swing Weight 3


Income 20 0.1
Attractiveness 80 0.4
Personality 100 0.5

Molly Niamh Uconn Girl


V(girl) using Weight 1 (Orig.) 75.6 78.9 11.1
V(girl) using tradeoff weights 60.4 62.2 34.5
V(girl) using equal weights 60 63.3 33.3
V(girl) using Weight 2 78.8 76.3 12.5
V(girl) using Weight 3 73 81 10

In all cases, Niamh is John’s future girlfriend with Molly as a close second. As John cares less and
less about personality, the better shot Molly has at being his boyfriend. I think it’s also safe to
say that UConn girl doesn’t have a snowball’s chance in hell. Sorry hun.
Michael Sandy Moore Homework #7 DADSS

You might also like