298 Exposure-Response Modeling: Methods and Practical Implementation
yi ~general(ll);
sp ecifies the log-likelihood function is g iven by variable ll and programmed
by the user . Note yi here is symbolic and has no impact on the model o r
on the calculation.
• PREDICT: This statement specifies what the user wants to predict based
on the fitted model. For example,
PREDICT pi out=pred;
PREDICT 1/(1+exp(beta0-Emax/((EC50/conci+1))) out=pred;
PREDICT 1/(1+exp(beta0-Emax/((EC50/100+1))) out=pred;
PREDICT 1/(1+exp(beta0+ui-Emax/((EC50/100+1))) out=pred;
The first one simply takes the pi as in the previous MODEL statement and
outputs it to dataset pred, together with variables in the input dataset.
The second one pre dic ts the probability without ui, hence it