pols 205 political science as a social science …3 review for final exam chris adolph (uw) music...
TRANSCRIPT
POLS 205Political Science as a Social Science
Music Preference Model Results
Christopher Adolph
University of Washington, Seattle
June 2, 2010
Chris Adolph (UW) Music Preferences June 2, 2010 1 / 44
1 Explanatory ModelsI ModelsI Out of Sample PerformanceI Substantive ResultsI Summary & Critique
2 Predictive ModelsI ModelsI Out of Sample Performance
3 Review for Final Exam
Chris Adolph (UW) Music Preferences June 2, 2010 2 / 44
Explanatory Models Models
−0.001to
0.11
0.11to
0.222
0.222to
0.333
0.333to
0.444
0.444to
0.556
0.556to
0.667
0.667to
0.778
0.778to
0.89
0.89to1
black
Res
iden
ceIn
stru
men
tR
ace
Reg
ion
Par
ents
Rel
igio
usP
arty
Mal
eD
ensi
ty
Team 6Team 5Team 7Team 3Team 2Team 1Team 4Team 9Team 8
Adolph PostAdolph Prior
9 Groups, and 9 differentexplanatory variables indifferent combinations
Which models best predictthe out of sample data?
Chris Adolph (UW) Music Preferences June 2, 2010 3 / 44
Explanatory Models Models
−0.001to
0.11
0.11to
0.222
0.222to
0.333
0.333to
0.444
0.444to
0.556
0.556to
0.667
0.667to
0.778
0.778to
0.89
0.89to1
black
Res
iden
ceIn
stru
men
tR
ace
Reg
ion
Par
ents
Rel
igio
usP
arty
Mal
eD
ensi
ty
Team 6Team 5Team 7Team 3Team 2Team 1Team 4Team 9Team 8
Adolph PostAdolph Prior
9 Groups, and 9 differentexplanatory variables indifferent combinations
Which models best predictthe out of sample data?
Chris Adolph (UW) Music Preferences June 2, 2010 3 / 44
Explanatory Models Out of Sample Performance
Out of sample tests
Amanda reserved 100 observations, randomly selected
How do our models estimated on the remaining training observations predictthe left-out data?
Use the model to fit the probability that each left-out individual prefers eachgenre
Call the most likely category the model prediction
Calculate the percent of cases correctly predicted
Chris Adolph (UW) Music Preferences June 2, 2010 4 / 44
Explanatory Models Out of Sample Performance
0.25 0.3 0.35 0.4 0.45
Proportion Correctly Predicted (out of sample)
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
●
●
●
●
●
●
●
●
●
●
●
Models ranked by out ofsample prediction
The dashed linecorresponds to predictingeveryone prefers rock, themodal category in thetraining sample
Chris Adolph (UW) Music Preferences June 2, 2010 5 / 44
Explanatory Models Out of Sample Performance
0.25 0.3 0.35 0.4 0.45
Proportion Correctly Predicted (out of sample)
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
●
●
●
●
●
●
●
●
●
●
●
Models ranked by out ofsample prediction
The dashed linecorresponds to predictingeveryone prefers rock, themodal category in thetraining sample
Chris Adolph (UW) Music Preferences June 2, 2010 5 / 44
Explanatory Models Out of Sample Performance
0.25 0.3 0.35 0.4 0.45
Proportion Correctly Predicted (out of sample)
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
●
●
●
●
●
●
●
●
●
●
●
How is our performance?
Chris Adolph (UW) Music Preferences June 2, 2010 6 / 44
Explanatory Models Out of Sample Performance
0.25 0.3 0.35 0.4 0.45
Proportion Correctly Predicted (out of sample)
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
●
●
●
●
●
●
●
●
●
●
●
One challenge: our testdataset differs more thanwe might expect
pop rap rocktrain 24% 30% 46%test 22 40 38
Chris Adolph (UW) Music Preferences June 2, 2010 7 / 44
Explanatory Models Substantive Results
Parents’ Preferences
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Parents Prefer Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 8 / 44
Explanatory Models Substantive Results
Parents’ Preferences
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Parents Prefer Rap
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 9 / 44
Explanatory Models Substantive Results
Parents’ Preferences
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Parents Prefer Pop
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 10 / 44
Explanatory Models Substantive Results
Density
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Urban
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 11 / 44
Explanatory Models Substantive Results
Density
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Rural
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 12 / 44
Explanatory Models Substantive Results
Density
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Suburban
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 13 / 44
Explanatory Models Substantive Results
Region
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Northwest
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 14 / 44
Explanatory Models Substantive Results
Region
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Other West
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 15 / 44
Explanatory Models Substantive Results
Region
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Midwest
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 16 / 44
Explanatory Models Substantive Results
Region
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
South
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 17 / 44
Explanatory Models Substantive Results
Region
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Outside US
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 18 / 44
Explanatory Models Substantive Results
Residence
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Dorms
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 19 / 44
Explanatory Models Substantive Results
Residence
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Greek Housing
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x1x1.5x2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 20 / 44
Explanatory Models Substantive Results
Residence
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Off−campus Housing
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 21 / 44
Explanatory Models Substantive Results
Party
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Democrat
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 22 / 44
Explanatory Models Substantive Results
Party
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Republican
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 23 / 44
Explanatory Models Substantive Results
Party
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Independent
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 24 / 44
Explanatory Models Substantive Results
Party
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Other Party
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 25 / 44
Explanatory Models Substantive Results
Religious Days
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Religious Days = 0
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 26 / 44
Explanatory Models Substantive Results
Religious Days
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Religious Days = 4
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 27 / 44
Explanatory Models Substantive Results
Religious Days
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Religious Days = 12
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1.5x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 28 / 44
Explanatory Models Substantive Results
Musical Instrument
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
No Instrument
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x1x1.5x2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 29 / 44
Explanatory Models Substantive Results
Musical Instrument
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Plays Instrument
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x1x1.5x2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 30 / 44
Explanatory Models Substantive Results
Gender
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Female
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 31 / 44
Explanatory Models Substantive Results
Gender
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Male
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 32 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
White
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 33 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Black
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 34 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Hisp
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 35 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Asian
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8
Team 9
Team 5
Adolph post
Adolph prior
Team 4
Team 1
Team 3
Team 6
Team 7
Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 36 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
Native American
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x1x1.5x2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 37 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
None given
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 38 / 44
Explanatory Models Substantive Results
Race
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rap
More than one
0 0.25 0.5 0.75 1
0.5x 1x 1.5x 2x
Rock
0 0.25 0.5 0.75 1
0.5x1x1.5x2x
Pop
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
● ● ●
Team 8Team 9Team 5
Adolph postAdolph prior
Team 4Team 1Team 3Team 6Team 7Team 2
Chris Adolph (UW) Music Preferences June 2, 2010 39 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Summary of Substantive Results
Mostly consistent but modest results across models
Probability of liking Rock rises for Whites, musicians, Males, and maybe otherwest & off campus
Probability of liking Rock falls for Blacks, Asians, Females, City-dwellers,Greek Housing, and possibly Hispanics & Republicans
Probabililty of liking Rap rises for Parents, Blacks, Asians, City-Dwellers, andpossibly Hispanics & Republicans
Probabililty of liking Rap falls for Whites & maybe Other West
Probability of liking Pop rises for Parents, Females, Asians, Greek Housing
Probability of liking Pop falls for Males, Musicians, Off-campus housing, andmaybe Blacks
Chris Adolph (UW) Music Preferences June 2, 2010 40 / 44
Explanatory Models Summary & Critique
Critique
Do you believe or doubt our results?
Which of the following may be problems, and why?
Selection biasOmitted variable biasEndoegneity / reverse causationMeasurement errorUnreliable measurementSpurrious correlation
Chris Adolph (UW) Music Preferences June 2, 2010 41 / 44
Explanatory Models Summary & Critique
Critique
Do you believe or doubt our results?
Which of the following may be problems, and why?
Selection biasOmitted variable biasEndoegneity / reverse causationMeasurement errorUnreliable measurementSpurrious correlation
Chris Adolph (UW) Music Preferences June 2, 2010 41 / 44
Predictive Models Models
−0.001to
0.11
0.11to
0.222
0.222to
0.333
0.333to
0.444
0.444to
0.556
0.556to
0.667
0.667to
0.778
0.778to
0.89
0.89to1
black
Spo
rts
TV
Par
ents
Bic
ycle
Inst
rum
ent
Res
iden
ceR
ace
Par
tyR
elig
ious
Reg
ion
Trus
tM
ovie
sG
roce
ryM
ale
Den
sity
Adolph PriorAdolph Post
Team 7Team 8Team 9Team 6Team 5Team 2Team 3Team 1Team 4
9 Groups, and 15 differentexplanatory variables indifferent combinations
Which models best predictthe out of sample data?
Chris Adolph (UW) Music Preferences June 2, 2010 42 / 44
Predictive Models Models
−0.001to
0.11
0.11to
0.222
0.222to
0.333
0.333to
0.444
0.444to
0.556
0.556to
0.667
0.667to
0.778
0.778to
0.89
0.89to1
black
Spo
rts
TV
Par
ents
Bic
ycle
Inst
rum
ent
Res
iden
ceR
ace
Par
tyR
elig
ious
Reg
ion
Trus
tM
ovie
sG
roce
ryM
ale
Den
sity
Adolph PriorAdolph Post
Team 7Team 8Team 9Team 6Team 5Team 2Team 3Team 1Team 4
9 Groups, and 15 differentexplanatory variables indifferent combinations
Which models best predictthe out of sample data?
Chris Adolph (UW) Music Preferences June 2, 2010 42 / 44
Predictive Models Out of Sample Performance
0.38 0.4 0.42 0.44 0.46 0.48
Proportion Correctly Predicted (out of sample)
Team 4
Team 3
Team 2
Adolph prior
Team 8
Adolph post
Team 1
Team 6
Team 5
Team 9
Team 7
●
●
●
●
●
●
●
●
●
●
●
Models ranked by out ofsample prediction
The dashed linecorresponds to predictingeveryone prefers rock, themodal category in thetraining sample
Chris Adolph (UW) Music Preferences June 2, 2010 43 / 44
Predictive Models Out of Sample Performance
0.38 0.4 0.42 0.44 0.46 0.48
Proportion Correctly Predicted (out of sample)
Team 4
Team 3
Team 2
Adolph prior
Team 8
Adolph post
Team 1
Team 6
Team 5
Team 9
Team 7
●
●
●
●
●
●
●
●
●
●
●
Models ranked by out ofsample prediction
The dashed linecorresponds to predictingeveryone prefers rock, themodal category in thetraining sample
Chris Adolph (UW) Music Preferences June 2, 2010 43 / 44
Predictive Models Out of Sample Performance
Congratulations!
Congratulations to Groups 2, 3, 4, 5, 8, and 9!
Each of these groups beat the predictions of one of my models
Each of these groups will receive 5% extra credit on the final exam
Chris Adolph (UW) Music Preferences June 2, 2010 44 / 44