Hi, I’ve got a question about the sequence command when used with a differential equation model. In the model below I’m using the sequence command to set the initial condition of each individual to a value of 169 which is the mean value for each of my individuals at time zero, but I have a measured value for each individual. What I would like to know is there a way I can set each individuals initial condition to their measured value either by entering each piece of data in or (preferably) reading it from the data? Thanks, Mike test(){ deriv(TuV = (lam0 - (k2*C))*TuV) cfMicro(A1, Ke, first = (Aa = Ka)) dosepoint(Aa) C = A1 / V sequence{TuV = 169} error(TuVEps = 1) observe(TuVObs = TuV * (1 + TuVEps)) stparm(Ka = tvKa) stparm(V = tvV) stparm(Ke = tvKe) stparm(lam0 = tvlam0 * exp(nlam0)) stparm(k2 = tvk2 * exp(nk2)) fixef(tvKa(freeze) = c(, 0.2, )) fixef(tvV(freeze) = c(, 0.5, )) fixef(tvKe(freeze) = c(, 2, )) fixef(tvlam0 = c(0, 0.0009, )) fixef(tvk2 = c(0, 0.0005, )) ranef(diag(nlam0, nk2) = c(1, 1)) }
Hi Mike, There are two options: 1. Use the value as a covariate. Put the initial condition in the dataset in a new column (e.g. “TuV0”). You will only need to fill in each individuals value at time = 0, leaving the rest blank. Put a statement “covariate(TuV0)” in your model and modify the sequence statement to “sequence{TuV = TuV0}”. This method assumes there is no error in your observation of the initial condition. 2. Let the algorithm solve for the starting point. Create a stparm for the initial condition and assign it in the sequence statement. fixef(tvTuV0 = 169) stparm(TuV0 = tvTuV0*exp(nTuV0)) sequence{TuV=TuV0} ranef(nTuV0 = 1) This method allows for observation error at the zero time. Happy modeling, -Jason
Hi Jason, I tried the second option and it works fine. Thanks for your help. Mike