Hi everyone, I am working on a combined model with LPV/RTV in pregnant patients and want to add gestational age or visit as a covariate. I have tried a few different things (using gestational age as a continuous covariate, visit as a categorical or occasional covariate, and also tried adding each visit as a different categorical covariate) When I try to run the code with the covariate having an effect on Clearance and Volume of LPV in QRPEM I get the following error message “Possible non linear covariate model is unsuitable for QRPEM”. However, when I try to run the model with FOCE-ELS I get -2LL values in the 2E+70 range. I also tried not mapping the covariate to any parameter to try and see if this would generate covariate plots but I wasn’t able to do that either. Are there any errors in my textual model or is there anyway to optimize my model to allow it to run in QRPEM mode? In one of the training manuals it suggests manually restructuring the model so that the non-linear covariate model is applied outside of the initial structural parameter definition rather than inside, what is the process of doing that? test(){ deriv(A1L = ((( - Cllpv * exp ( - coef * (Cmrtv))) / Vlpv) * A1L) + (Kalpv * Aalpv)) urinecpt(A0 = (Cllpv * Clpv)) deriv(Aalpv = - (Aalpv * Kalpv)) deriv(A2R = (Aartv * Kartv)- (Cllrtv * Crtv)) deriv(Aartv = - (Aartv * Kartv)) urinecpt(A0 = (Cllrtv * Crtv)) Clpv = A1L / Vlpv dosepoint(Aalpv, tlag = (Tlaglpv)) error(CEpslpv = 0.449944) observe(CObslpv = Clpv * (1 + CEpslpv)) Crtv = A2R / Vrtv Cmrtv = A2R / Vrtv dosepoint(Aartv, tlag = (Tlagrtv)) error(CEpsrtv = 0.579401) observe(CObsrtv = Crtv * (1 + CEpsrtv)) error(CEpsmrtv = 144.67) observe(CObsmrtv = Cmrtv + CEpsmrtv) stparm(Vlpv = tvVlpv * exp(nVlpv + nVlpvocc_1*(occ==1) + nVlpvocc_2*(occ==2)+ nVlpvocc_3*(occ==3)+ nVlpvocc_4*(occ==4)) stparm(Cllpv = tvCllpv * exp(nCllpv + nCllpvocc_1*(occ==1) + nCllpvocc_2*(occ==2) + nCllpvocc_3*(occ==3) + nCllpvocc_4*(occ==4)) stparm(Kalpv = tvKalpv * exp(nKalpv)) stparm(Vrtv = tvVrtv * exp(nVrtv)) stparm(Kartv = tvKartv * exp(nKartv)) stparm(Cllrtv = tvCllrtv * exp(nCllrtv)) stparm(Tlaglpv = tvTlaglpv * exp(nTlaglpv)) stparm(Tlagrtv = tvTlagrtv * exp(nTlagrtv)) stparm(coef = tvcoef * exp(ncoef)) covariate(occ) fixef(tvVlpv = c(, 116925, )) fixef(tvCllpv = c(, 31941.7, )) fixef(tvKalpv = c(, 0.148752, )) fixef(tvVrtv = c(, 905439, )) fixef(tvKartv = c(, 0.40522, )) fixef(tvCllrtv = c(, 150284, )) fixef(tvTlaglpv = c(, 0.861153, )) fixef(tvTlagrtv = c(, 1.3694, )) fixef(tvcoef = c(, 0.000898966, )) ranef (diag(nVlpvocc_1) = c (1), same (nVlpvocc_2), same (nVlpvocc_3), same (nVlpvocc_4)) ranef (diag(nCllpvocc_1) = c (0.2), same (nCllpvocc_2), same (nCllpvocc_3), same (nCllpvocc_4)) ranef(diag(nKalpv, nKartv, nVrtv, nCllpv, nCllrtv, nTlaglpv, nTlagrtv, ncoef, nVlpv) = c(0.00028215382, 0.17697475, 0.099161057, 0.012838487, 0.11332717, 2.1024374E-06, 0.014681594, 0.0045704313, 1.1434849E-06)) } Thanks, Izna
testing digest mode
Dear Izna
In 1.3 version, inter-occasion is not supported with QRPEM. In 1.4, all the code you wrote should work as there is no non linear covariate but only use of inter-occasional covariate.
I believe that your model is way over-parametrized which caused the likelihood to diverge to infinite. Can you attach your data set and project? This will enable me to help you more. I believe you should use a model building approach and step by step increase the complexity of your model if it is supported by statistical tools. I can help you with that but only if I have the data set and project (the project should be enough as it would have the data set).
best
Serge