glmfit with multivaria​nt('binomi​al','link'​,'probit')

Somebody helped me with a code about probit regression to test whether one factor may affect the mu and threshold between two curves，but I don't quite understand some details about the code. Here is one example:

X1=[-8 -3.2 -1.28 -0.51 0 0.51 1.28 3.20 8 -8 -3.2 -1.28 -0.51 0 0.51 1.28 3.20 8];

X2=[0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1];

Y1=[0 0 3 4 5 7 9 10 10 0 4 3 7 6 8 10 9 10];

Y2=[10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10]

[b, dev, stats] = glmfit([ X1 X2 X1.* X2],[Y1 Y2],'binomial','link','probit');

%Case I

bias(1)= (norminv(0.5)-b(1))/b(2);% 50 pct PD threshold

threshold(1) = abs( norminv(0.84)/b(2) ); % 84% gaussian threshold

%Case2 bias(2) = (norminv(0.5)-(b(1) + b(3)))/(b(2) + b(4)); threshold(2) = abs( norminv(0.84)/(b(2)+b(4)) ); % p value P_bias = stats.p(3); % P value for bias P_slope = stats.p(4); % P value for threshold

I don't quite understand why bias(1),threshold(1),bias(2),threshold(2) can be formulated like that？And why stats.p(3) and stats.p(4) can imply the bias and threshold difference between two factors?