Here is an example of the incorrect use of INTEGER. Suppose the DV is non continuous (e.g., takes values 0, 1, 2, ...). Suppose F is continuous. One could make it take values 0, 1, 2 ... with this code:
$ERROR Y=INT(F) + ERR(1) But now Y is discontinuous. A change in a theta should produce a change in Y. With this code it may not change Y at all. (E.g., if F changes from 2.1 to 2.2, Y is still 2). The Estimation step will fail. I'm not sure what the error message from NONMEM would be. This is why there are special techniques (e.g., logistic regression) for dealing with odd-type data. Not only Y, but also its first and second derivatives must be continuous. I think that INTEGER can be used safely in a model with ordinary data (continuous Y and DV). What matters is that values of TIME at which there is a discontinuity must *not* be values of TIME at which the objective function is evaluated, that is, must not coincide with Event record times. On Tue, 31 May 2011 13:52 -0400, "Li Li" <lili.uf2...@gmail.com> wrote: Hi, I have two questions about warning 68: (WARNING 68) THE INTEGER FUNCTION IS BEING USED OUTSIDE OF A SIMULATION BLOCK. IF THE INTEGER VALUE AFFECTS THE VALUE OF THE OBJECTIVE FUNCTION, THEN AN ERROR WILL PROBABLY OCCUR. First question: In what kind of cases the Integer value will affect the value of the objective function? Second question: If the integer value affects the value of objective function, then what kind of errors will occur? I understand that there might be multiple situations/explanations for this warning. I just want to get the general concepts. Thank you. best regards, Li li -- Alison Boeckmann alisonboeckm...@fastmail.fm
