Module 2: Preference Modeling w. Notes

data( list="bandpref", package = "MPsychoR")
bandpref

Btm_bandpref <-BradleyTerry2::BTm(outcome=cbind(Win1, Win2), player1 = Band1, player2 = Band2, data = bandpref)
BTabilities_bandpref <-BradleyTerry2::BTabilities(model = Btm_bandpref)

BTabilities_bandpref <-BTabilities_bandpref[order(BTabilities_bandpref[,"ability"]),]
BTabilities_bandpref

              ability       s.e.
Scorpions -0.31178892 0.09099633
Emperor   -0.02434454 0.09008490
Slayer     0.00000000 0.00000000
Death      0.19891105 0.09026853
Rush       0.37987618 0.09098862

#Here the Slayer is chosen to be the reference level. It does not esitmate the standard error. All others are in reference to slayer. #Rusher is the most favored overall. Scorpion is lesast favored.

v_alpha <-exp(BTabilities_bandpref[,"ability"])
v_alpha

Scorpions   Emperor    Slayer     Death      Rush 
0.7321361 0.9759494 1.0000000 1.2200734 1.4621035 

#Here are the odds that Rush is prefered at 1.20 to Death.

v_alpha["Rush"]/v_alpha["Death"]

    Rush 
1.198373 

#But, why is one prefered to the other? Bradley Terry trees uses trees to see the relationship between covariates and preferences. Tree models try to find fine-grained subgroup combinations that influence the preferences.

data(list = "Topmodel2007", package = "psychotree")
head(Topmodel2007)

mob <- partykit::mob
bttree_Topmodel2007 <-psychotree::bttree(formula=preference ~ age + gender + q1 + q2 + q3, data = Topmodel2007)

plot(x=bttree_Topmodel2007)

#If we were to split all of the subject between two groups, we should split on age on 52. Brb is preferred. Second name is not as liked, then the others. Q2 is about TV. Yes likes Hannah. No, then is split between gender. Males and females differ.