duplicate observed dose predicted

[R-garreau]

Describe the bug

When optimizing posterior prediction with bestdose it does duplicate the firt dose (the observed one).
See below.
Of Note nothing has changed in BDcore and all test are still passing, and this behavior was not here before, which means the bug probably lives in BDmichael branch

Input for prediction

target
   id evid     time dur dose addl ii input   out outeq c0  c1 c2 c3
1   1    1  0.00000   0  150    0  0     1    NA    NA  0 0.2  0  0
2   1    0  2.00000  NA   NA   NA NA    NA 0.759     1  0 0.2  0  0
3   1    0  4.00000  NA   NA   NA NA    NA 0.384     1  0 0.2  0  0
4   1    0  6.00000  NA   NA   NA NA    NA 0.194     1  0 0.2  0  0
5   1    1 12.00000   0   75    0  0     1    NA    NA  0 0.2  0  0
6   1    0 14.00000  NA   NA   NA NA    NA 0.392     1  0 0.2  0  0
7   1    0 16.00000  NA   NA   NA NA    NA 0.198     1  0 0.2  0  0
8   1    0 18.00000  NA   NA   NA NA    NA 0.100     1  0 0.2  0  0
9   1    1 24.00000   0    0   NA NA     1    NA    NA  0 0.2  0  0
10  1    0 36.00000  NA   NA   NA NA    NA 0.250     1  0 0.2  0  0
11  1    1 36.01667   0    0   NA NA     1    NA    NA  0 0.2  0  0
12  1    0 48.00000  NA   NA   NA NA    NA 0.250     1  0 0.2  0  0
13  1    1 48.01667   0    0   NA NA     1    NA    NA  0 0.2  0  0
14  1    0 60.00000  NA   NA   NA NA    NA 0.250     1  0 0.2  0  0
15  1    1 60.01667   0    0   NA NA     1    NA    NA  0 0.2  0  0
16  1    0 72.00000  NA   NA   NA NA    NA 0.250     1  0 0.2  0  0

r$> dose_range
$min
[1] 0

$max
[1] 1e+05


r$> start
[1] 24

r$> options$prior_weight
[1] 0

csv file generated

id, evid, time            , dur, dose, addl, ii, input, out  , outeq, c0, c1 , c2, c3
 1,    1,  0              , 0  , 150 , 0   , 0 , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0,  2              , .  , .   , .   , . , .    , 0.759, 1    ,  0, 0.2,  0,  0
 1,    0,  4              , .  , .   , .   , . , .    , 0.384, 1    ,  0, 0.2,  0,  0
 1,    0,  6              , .  , .   , .   , . , .    , 0.194, 1    ,  0, 0.2,  0,  0
 1,    1, 12              , 0  , 75  , 0   , 0 , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0, 14              , .  , .   , .   , . , .    , 0.392, 1    ,  0, 0.2,  0,  0
 1,    0, 16              , .  , .   , .   , . , .    , 0.198, 1    ,  0, 0.2,  0,  0
 1,    0, 18              , .  , .   , .   , . , .    , 0.1  , 1    ,  0, 0.2,  0,  0
 1,    1, 24              , 0  , 0   , .   , . , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0, 36              , .  , .   , .   , . , .    , 0.25 , 1    ,  0, 0.2,  0,  0
 1,    1, 36.0166666666667, 0  , 0   , .   , . , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0, 48              , .  , .   , .   , . , .    , 0.25 , 1    ,  0, 0.2,  0,  0
 1,    1, 48.0166666666667, 0  , 0   , .   , . , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0, 60              , .  , .   , .   , . , .    , 0.25 , 1    ,  0, 0.2,  0,  0
 1,    1, 60.0166666666667, 0  , 0   , .   , . , 1    , .    , .    ,  0, 0.2,  0,  0
 1,    0, 72              , .  , .   , .   , . , .    , 0.25 , 1    ,  0, 0.2,  0,  0

output

BestDose Optimization Results
==============================

Optimal doses: [150.00, 75.00, 150.00, 75.00, 1487.78, 1455.69, 1455.68, 1455.74] mg
Objective function: 0.0000007711
ln(Objective): -14.0755
Method: Posterior
Status: Converged
Prior weight: 0.00
Start offset (h): 24.00

Number of predictions: 10

Expected behavior

It should return the following value :
Optimal doses: [150.00, 75.00, 1487.78, 1455.69, 1455.68, 1455.74]

[mhovd]

I notice that only doses with addl and ii are duplicated.
I wonder if this could be related to the expansion data that was just added to Pmetrics.

To check, could you try again and set addl and ii to zero for the other doses as well?

[R-garreau]

I have the same results with :

• all addl and ii set to NA
• all addl and ii set to 0