ERROR: TaskFailedException: AssertionError: cur_ev.time == t

1

Can someone help me to solve this

VCWTresult1  = fit(VCWTmodel1,VCWTDF1, VCWTparam1, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
Iter     Function value   Gradient norm
     0              Inf              NaN
 * time: 0.0
ERROR: TaskFailedException:
AssertionError: cur_ev.time == t
Stacktrace:
 [1] solve(::AnalyticalPKPDProblem{LabelledArrays.SLArray{Tuple{4},ForwardDiff.Dual{ForwardDiff.Tag{Pumas.var"#318#319"{Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{,Tuple{}}},PumasModel{ParamSet{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},PDiagDomain{PDMats.PDiagMat{Float64,Array{Float64,1}}},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64}}}},var"#1269#1295",var"#1270#1296",var"#1272#1298",Pumas.LinearODE,var"#1273#1299",var"#1284#1310"},Subject{NamedTuple{(:DV,),Tuple{Array{Union{Missing, Float64},1}}},Pumas.ConstantCovar{NamedTuple{(:WT, :INITWT, :DFWT),Tuple{Float64,Float64,Float64}}},Array{Pumas.Event{Float64,Float64,Float64,Float64,Float64,Float64,Nothing,Int64},1},Array{Float64,1}},NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{ForwardDiff.Dual{ForwardDiff.Tag{Pumas.var"#295#296"{Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{,Tuple{}}},PumasModel{ParamSet{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},PDiagDomain{PDMats.PDiagMat{Float64,Array{Float64,1}}},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64}}}},var"#1269#1295",var"#1270#1296",var"#1272#1298",Pumas.LinearODE,var"#1273#1299",var"#1284#1310"},Subject{NamedTuple{(:DV,),Tuple{Array{Union{Missing, Float64},1}}},Pumas.ConstantCovar{NamedTuple{(:WT, :INITWT, :DFWT),Tuple{Float64,Float64,Float64}}},Array{Pumas.Event{Float64,Float64,Float64,Float64,Float64,Float64,Nothing,Int64},1},Array{Float64,1}},Array{Float64,1},Pumas.FOCEI,TransformVariables.TransformTuple{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},Pumas.PDiagTransform,TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64}}}}},Float64},Float64,1},ForwardDiff.Dual{ForwardDiff.Tag{Pumas.var"#295#296"{Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{,Tuple{}}},PumasModel{ParamSet{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},PDiagDomain{PDMats.PDiagMat{Float64,Array{Float64,1}}},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64}}}},var"#1269#1295",var"#1270#1296",var"#1272#1298",Pumas.LinearODE,var"#1273#1299",var"#1284#1310"},Subject{NamedTuple{(:DV,),Tuple{Array{Union{Missing, Float64},1}}},Pumas.ConstantCovar{NamedTuple{(:WT, :INITWT, :DFWT),Tuple{Float64,Float64,Float64}}},Array{Pumas.Event{Float64,Float64,Float64,Float64,Float64,Float64,Nothing,Int64},1},Array{Float64,1}},Array{Float64,1},Pumas.FOCEI,TransformVariables.TransformTuple{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ, :tvpintwt, :tvpdfwt),Tuple{TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},Pumas.PDiagTransform,TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64},TransformVariables.ScaledShiftedLogistic{Float64}}}}},Float64},Float64,1},ForwardDiff.Dual{ForwardDiff.Tag{Pumas

my model is

VCWTmodel1 = @model begin
  @param begin
    tvka  ∈ RealDomain(lower=0.1, upper = 5)
    tvk23 ∈ RealDomain(lower=0.1, upper = 20 )
    tvQ   ∈ RealDomain(lower=0.1, upper = 150 )
    tvvc  ∈ RealDomain(lower=0.1, upper = 150)
    tvvp  ∈ RealDomain(lower=0.1, upper = 150)
    tvCL  ∈ RealDomain(lower=0.1 ,upper =15)
    tvbio1 ∈ RealDomain(lower=0.1, upper =1)
    Ω     ∈ PDiagDomain(7) # CL = 0.23,Vc = 0.72,KA = 3.37,vp = 0.04,k23 = 0.09
    σ     ∈ RealDomain(lower=0.01,upper = 1)
    tvpintwt ∈ RealDomain(lower=0.1, upper = 10)
    tvpdfwt  ∈ RealDomain(lower = 0.1,upper =10)
  end
  @random begin
    η ~ MvNormal(Ω)
  end
  @covariates WT INITWT DFWT
  @pre begin
    ka    = tvka  * exp(η[1])
    k23   = tvk23 * exp(η[2])
    Q     = tvQ  * exp(η[3])
    pintwt = tvpintwt
    pdfwt  = tvpdfwt
    vc    = tvvc* (1 + pintwt*(INITWT-14.3) + pdfwt*DFWT) * exp(η[4])
    vp    = tvvp * exp(η[5])
    CL    = tvCL  * exp(η[6])
    bioav = (Depot1 = tvbio1* exp(η[7]))

    A = @SMatrix([-ka    0      0           0
                   0    -k23    0           0
                  ka    k23 -(CL+Q)/vc  Q/vp
                  0       0    Q/vc       -Q/vp  ])
  end
  @init begin
    Depot1 = 0.0
    Prodrug = 0.0
    Central = 0.0
    Peripheral = 0.0
  end
  @dynamics LinearODE
  @derived begin
    conc := @. Central/vc
    DV ~ @. Normal(conc, σ*abs(conc))
  end
end
2

model has time varying covariate. INITWT is initial weight, WT = current weight, DFWT = difference from initial weight to current weight

3

It looks like the value of (minus) the log-likelihood is Inf in the starting values. Could you please confirm by just running loglikelihood(model, population, initial_parameters, your_estimation_method)? If that is the case then you’d have to find some other starting values that give a finite log-likelihood.

4 

julia> VCWTll =  loglikelihood(VCWTmodel1,VCWTDF1, VCWTparam1, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
NaN

is that infinity?

5

No that’s a NaN (no not a number). They show up in cases like

julia> 0/0
NaN

julia> Inf - Inf
NaN

My best guess is that one of the volumes become zero. Maybe try to call Pumas.findinfluential instead of loglikelihood. That will show you a list of the subjects that contribute the most to the loglikelihood.

6 

julia> Pumas.findinfluential(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
5-element Array{Tuple{String,Float64},1}:
 ("41", 126.79140281385544)
 ("172", 46.0780610366132)
 ("44", 44.02736630754179)
 ("10", 41.0442973131424)
 ("171", 37.20470753972811)
7

Do i need to remove them?

8

this result is after removing BSV on both volumes

9

I suppose that the value of loglikelihood is then no longer NaN or Inf after this change. Is that the case?

10

even after removing BSV on both volumes, loglikelihood is NaN .
I tried to find influential after that.
result is

julia> Pumas.findinfluential(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
5-element Array{Tuple{String,Float64},1}:
 ("41", 126.79140281385544)
 ("172", 46.0780610366132)
 ("44", 44.02736630754179)
 ("10", 41.0442973131424)
 ("171", 37.20470753972811)

I checked observations of those ID’s Manually, They were very different in same ID.

ex: ID 10 has conc of 1.9mg/L after 5 hours of 200 mg of dose, same id has 1.9 mg/L after 10 hours of 200mg dose administration on other day. These 5 ID’s have that kind of data.

11

Could you try pasting exactly this into your session loglikelihood(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads()) and let me know the result

12

Hi @andreasnoack, Here is result

VCWTmodel1 = @model begin
  @param begin
    tvka  ∈ RealDomain(lower=0.1, upper = 5)
    tvk23 ∈ RealDomain(lower=0.1, upper = 20 )
    tvQ   ∈ RealDomain(lower=0.1, upper = 250 )
    tvvc  ∈ RealDomain(lower=0.1, upper = 250)
    tvvp  ∈ RealDomain(lower=0.5, upper = 250)
    tvCL  ∈ RealDomain(lower=0.1 ,upper =15)
    tvbio1 ∈ RealDomain(lower=0.1, upper =1)
    Ω     ∈ PDiagDomain(5) # CL = 0.23,Vc = 0.72,KA = 3.37,vp = 0.04,k23 = 0.09
    σ     ∈ RealDomain(lower=0.1,upper = 1)
    #tvpintwt ∈ RealDomain(lower=0.001, upper = 100)
    #tvpdfwt  ∈ RealDomain(lower = 0.001,upper =100)
  end
  @random begin
    η ~ MvNormal(Ω)
  end
  @covariates WT INITWT DFWT
  @pre begin
    ka    = tvka  * exp(η[1])
    k23   = tvk23 * exp(η[2])
    Q     = tvQ  * exp(η[3])
  #  pintwt = tvpintwt
  #  pdfwt  = tvpdfwt
  #  vc    = tvvc * (1 + pintwt*(INITWT-14.3) + pdfwt*DFWT)
    vc    = tvvc * (WT/14.3)
    vp    = tvvp
    CL    = tvCL  * exp(η[4])
    bioav = (Depot1 = tvbio1* exp(η[5]))

    A = @SMatrix([-ka    0      0           0
                   0    -k23    0           0
                  ka    k23 -(CL+Q)/vc  Q/vp
                  0       0    Q/vc       -Q/vp  ])
  end
  @init begin
    Depot1 = 0.0
    Prodrug = 0.0
    Central = 0.0
    Peripheral = 0.0
  end
  @dynamics LinearODE
  @derived begin
    conc := @. Central/vc
    DV ~ @. Normal(conc, σ*abs(conc))
  end
end


VCWTparam1dup = (tvka   = 1.9,
              tvk23  = 0.18,
              tvQ    = 5.23,
              tvvc   = 2.27,
              tvvp   = 44.86,
              tvCL   = 0.8,
              tvbio1 = 0.2,
              Ω      = Diagonal([0.05,0.004,0.01,0.04,0.08]),
              σ       = 0.45)

VCWTDF1 = read_pumas(phen1, observations = [:DV], id = :ID, time = :TIME, amt = :AMT, rate = :RATE, cmt = :CMT,
                covariates = [:WT,:INITWT,:DFWT])


fit(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads())

fit resulted in below error

ERROR: InexactError: trunc(Int64, Inf)
Stacktrace:
 [1] trunc at .\float.jl:703 [inlined]
 [2] ceil at .\float.jl:365 [inlined]
 [3] ceil at C:\Users\julia\.julia\packages\ForwardDiff\sdToQ\src\dual.jl:265 [inlined]
 [4] _exp(::Size{(4, 4)}, ::SArray{Tuple{4,4},ForwardDiff.Dual{ForwardDiff.Tag{Pumas.var"#318#319"{Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{,Tuple{}}},PumasModel{ParamSet{NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ),Tuple{RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},RealDomain{Float64,Int64,Float64},PDiagDomain{PDMats.PDiagMat{Float64,Array{Float64,1}}},RealDomain{Float64,Int64,Float64}}}},var"#53#75",var"#54#76",var"#56#78",Pumas.LinearODE,var"#57#79",var"#66#88"},Subject{NamedTuple{(:DV,),Tuple{Array{Union{Missing, Float64},1}}},Pumas.ConstantInterpolationStructArray{Array{Float64,1},StructArrays.StructArray{NamedTuple{(:WT, :INITWT, :DFWT),Tuple{Float64,Float64,Float64}},1,NamedTuple{(:WT, :INITWT, :DFWT),Tuple{Array{Float64,1},Array{Float64,1},Array{Float64,1}}},Int64},Symbol},Array{Pumas.Event{Float64,Float64,Float64,Float64,Float64,Float64,Nothing,Int64},1},Array{Float64,1}},NamedTuple{(:tvka, :tvk23, :tvQ, :tvvc, :tvvp, :tvCL, :tvbio1, :Ω, :σ),Tuple{Float64,Float64,Float64,Float64,Float64,Float64,Float64,PDMats.PDiagMat{Float64,Array{Float64,1}},Float64}}},Float64},Float64,5},2,16}) at C:\Users\julia\.julia\packages\StaticArrays\LJQEe\src\expm.jl:97

then i tried loglikelihood and influential codes

julia> loglikelihood(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
-870.0609562653066

julia> Pumas.findinfluential(VCWTmodel1,VCWTDF1, VCWTparam1dup, Pumas.FOCEI(),ensemblealg = EnsembleThreads())
5-element Array{Tuple{String,Float64},1}:
 ("41", 126.79140281385544)
 ("172", 46.0780610366132)
 ("44", 44.02736630754179)
 ("10", 41.0442973131424)
 ("171", 37.20470753972811)

It did not give NAN for loglikelihood after removing BSV on both volumes .
and fit is not yet successful.