Chapter 6: basic lever-pressing example

We will perform the calculations based on observing 3 failed (no-food) lever-presses and 2 successful lever-presses in two ways. First, we will iteratively compute the evidence based on the binomial probabilities for single observations of successful and unsuccessful lever-presses (definitions given in the main text), and then we will arrive at the same result using an all-at-once calculation based on the binomial probabilities describing the full dataset under each of the two models.

%%% press-by-press calculation

dat=[0 0 1 0 1];

dEpos=10*log10(3/1); %evidence change (high-rate hypothesis) following successful press

dEneg=10*log10(7/9); %evidence change (high-rate hypothesis) following unsuccessful press

EV=0;

for i=1:length(dat),

if dat(i)==1, EV=EV+dEpos;

else EV=EV+dEneg; end, end

%%% all-at-once calculation

EV(2)=sum(dat==1)*dEpos+sum(dat==0)*dEneg;

%%% alternate all-at-once calculation

rhoH=.3; rhoL=.1;

Lh=bpdf(sum(dat),length(dat),rhoH);

Ll=bpdf(sum(dat),length(dat),rhoL);

EV(3)=10*log10(Lh/Ll);

%%% compare the three calculations

isnear(EV(1),EV(2))

isnear(EV(2),EV(3))