Re: Interpreting the coefficient in a GLS binomial model

With 9 jurors there will be BIN(9,2), or 36 different pairs. The data looks like:
j_pair,count_C,count_O,index

were "j_pair" is a pair of jurors, "count_C" the number of concurring votes, "count_O" the number of dissenting votes, "index" the distance. Concurring votes are more probable then the dissenting votes because the Yes-votes are positively correlated. It is simple to show that No-No votes are then also positively correlated, and Yes-No, No-Yes votes are negatively correlated. Moreover,
corr{yes_yes}=corr(no-no)=-corr(yes,no)=-corr(no,yes)

Serguei
.