how math can help you choose the right college dr. geoff turner
TRANSCRIPT
How Math Can Help You Choose the Right College
Dr. Geoff Turner
What Today’s Talk WON’T Do
Sadly, I can’t tell you what your ideal college is.
Nor can I provide an algorithm for finding your ideal college. (But then again, no one can.)
Not how psychologists typically use math.
What Today’s Talk Will (Hopefully) Do
• Let you know how psychologists make decisions and think about how people make decisions (using math)
• Suggest ideas to help you in your search for a college
• Available on the web– web.simmons.edu/~turnerg/choice/choice.ppt
Math in Psychology
Statistical significance:
A difference is not always a difference!
Math in Psychology
Statistical significance:
A difference is not always a difference!
Significant Differences
Significant Differences
14.01
15.46
14.01
15.46
10.00
12.00
14.00
16.00
18.00
20.00
22.00
0 1 2 3 4
Significant Differences
14.01
15.46
14.01
15.46
10.00
12.00
14.00
16.00
18.00
20.00
22.00
0 1 2 3 4
Significant Differences
Detecting Deception
Do “lie detectors” really work?
Polygraph Test
0
10
20
30
40
50
60
70
80
90
Guilty people judged guilty Innocent people judged guilty
Percent
It works!
Polygraph Test
0
10
20
30
40
50
60
70
80
90
Guilty people judged guilty Innocent people judged guilty
Percent
Oops!
How Psychologists Judge Differences
The Mathematical Models
RH = P(“remember”|Old) ⎟⎠
⎞⎜⎝
⎛Φ⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛−
′+′
′′Φ=
tC
Cdd
dd
tr
o
yx
yx
22
21
RF =P(“remem ber”|New) ( )⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛+
′+′
′−′Φ−Φ= r
yx
yxo C
dd
ddC
22
22
F = P(“old”|New) ( )oC−Φ=
H = P(“old”|Old)⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛−
′+′
′′Φ= o
yx
yx Cdd
dd
t 22
21
Choices, Choices
1. People don’t like choosing (deciding)
2. People have surprisingly little insight into their own thought processes.
3. People’s choices are remarkably inconsistent over time, even under apparently identical conditions.
4. Frequently, our choices are not optimal. They’re irrational.
What’s for dinner?
If offered the choice between beef and chicken, you might choose beef.
Beef ≻ Chicken
What’s for dinner?
If offered the choice between chicken and fish, you might choose chicken.
Chicken ≻ Fish
What’s for dinner?
If offered the choice between fish and beef, then, of course you would choose …
Fish ≻ Beef ?
Beef ≻ Fish ?
How is this possible?
• Beef ≻ Chicken
• Chicken ≻ Fish
• Fish ≻ Beef
What’s for dinner?
Problem with Decision Models
Intransitivity (the paper, rock, scissors problem)
The transitive property
If a > b, and b > c, then a > c
but…
Back to Dinner
• Beef ≻ Chicken (Taste)
• Chicken ≻ Fish (Taste)
• Fish ≻ Beef (Health)
1 decision, but 2 dimensions
Two Dimensions
Beef
Chicken
Fish0
5
10
15
20
25
0 5 10 15 20 25
Health
Taste
Transitivity Within Dimension
Same Potential Issue With College Choice
• Simmons ≻ BU
• BU ≻ Northeastern
• Northeastern ≻ Simmons
Most Real-World Decisions Are Like This
What Can We Do?
(besides flip a coin)
Multi-Dimensional Scaling
Discover relationships from comparisons:
LA-NY > LA-Denver
LA-Atlanta > Seattle-SF
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
Multi-Dimensional Scaling
Repertory Grid (Kelly, 1955)
• Gather a set of schools you think you might be interested in
• Enumerate all possible triples
# Schools # Triples
3 1
4 3
5 6
6 10
Next
• Split each triple based on “feel” or intuition - whatever comes naturally.
• Define the way members of the pair are similar (and why they make a pair) and how the third is different.
• Example: BU, Simmons, Northeastern
BU, NEU vs Simmons Large vs. Small
Repertory Grid (Kelly, 1955)
Construct a matrix of comparisons:
BU Wheelock Simmons Harvard NEU
Size: Large vs Small
1 0 1
Co-ed vs. women
0 0 1 0 0
Focus: Research vs
Teaching1 0 0 1 1
Faculty: Brilliant vs Ordinary
0 0 1 1 0
Good food vs. bad
0 0 1 1 1
Repertory Grid (Kelly, 1955)
Construct a matrix of comparisons:
BU Wheelock Simmons Harvard NEU
Size: Large vs Small
1 0 0 1 1
Co-ed vs. women
0 0 1 0 0
Focus: Research vs
Teaching1 0 0 1 1
Faculty: Brilliant vs Ordinary
0 0 1 1 0
Good food vs. bad
0 0 1 1 1
Repertory Grid (Kelly, 1955)
• Sort the Matrix by Element (School) putting similar together
BU NEU Harvard Wheelock Simmons
Size: Large vs Small
1 1 1 0 0
Co-ed vs. women
0 0 0 0 1
Focus: Research vs
Teaching1 1 1 0 0
Faculty: Brilliant vs Ordinary
0 0 1 0 1
Good food vs. bad
0 1 1 0 1
Repertory Grid (Kelly, 1955)
• Sort the Matrix by Construct (Attribute) putting similar together
BU NEU Harvard Wheelock Simmons
Size: Large vs Small
1 1 1 0 0
Focus: Research vs
Teaching1 1 1 0 0
Co-ed vs. women
0 0 0 0 1
Faculty: Brilliant vs Ordinary
0 0 1 0 1
Good food vs. bad
0 1 1 0 1
Results
• BU and NEU are nearly identical; further examination may be necessary.
• Size and Research are equated - should they be or is this a bias? Double counting this influence?
• Which constructs are most important to you?
BU NEU Harvard Wheelock Simmons
Size: Large vs Small
1 1 1 0 0
Focus: Research vs
Teaching1 1 1 0 0
Co-Ed vs. Women
1 0 0 1 1
Faculty: Brilliant vs Ordinary
0 0 1 0 1
Good food vs bad
0 1 1 0 1
What College Is Best For You?
Simmons!
What College Is Best For You?