search

UUPharmacometrics / piraid

An aid for development and diagnosis of pharmacometric IRT models
GNU Lesser General Public License v3.0
10 stars 2 forks source link

Bug with add_lv_model #36

Open pmpharm opened 4 years ago

pmpharm commented 4 years ago

Tried to add two complementary LV models as follows:

m5 <- reset_lv_models(m4)
m6 <- add_lv_model(m5, "linear", c(2,4,8,10,14,19,25,29,34,36,41,47,50,52,57))
m6 <- add_lv_model(m6, "linear", setdiff(1:58,c(2,4,8,10,14,19,25,29,34,36,41,47,50,52,57)))

When run got the NM error: RANDOM VARIABLE IS DEFINED IN A NESTED IF STRUCTURE. Fixed the problem by modified the output code from piraid from: 

IF (PSI_MODEL.EQ.1) THEN
    BASE=THETA(233)+ETA(1)
    SLP=THETA(234)+ETA(2)
    PSI=BASE+SLP*TIME
ELSE IF (PSI_MODEL.EQ.2) THEN
    BASE=THETA(235)+ETA(3)
    SLP=THETA(236)+ETA(4)
    PSI=BASE+SLP*TIME
ENDIF

to 

IF (PSI_MODEL.EQ.1) THEN
    BASE=THETA(233)+ETA(1)
    SLP=THETA(234)+ETA(2)
    PSI=BASE+SLP*TIME
ELSE 
    BASE=THETA(235)+ETA(3)
    SLP=THETA(236)+ETA(4)
    PSI=BASE+SLP*TIME
ENDIF

Models run without NM error after.

pmpharm commented 4 years ago

Same in different, gave the same NM error:

> m5 <- reset_lv_models(m4)
> m11 <- add_lv_model(m5, "linear", c(2,4,8,10,14,19,25,29,34,36,41,47,50,52,57))
> m11 <- add_lv_model(m11, "linear", c(3,5,12,16,20,23,26,30,32,37,40,42,43,53,55,58))
> m11 <- add_lv_model(m11, "linear", c(6,11,17,27,35,45,49))
> m11 <- add_lv_model(m11, "linear", c(1,7,13,15,18,21,24,28,31,38,39,44,48,51,54,56))
> m11 <- add_lv_model(m11, "linear", c(9,22,33,46))
> save_model_code(m11, "run11.mod")

Fixed by changing:

> IF (PSI_MODEL.EQ.1) THEN
>     BASE=THETA(233)+ETA(1)
>     SLP=THETA(234)+ETA(2)
>     PSI=BASE+SLP*TIME
> ELSE IF (PSI_MODEL.EQ.2) THEN
>     BASE=THETA(235)+ETA(3)
>     SLP=THETA(236)+ETA(4)
>     PSI=BASE+SLP*TIME
> ELSE IF (PSI_MODEL.EQ.3) THEN
>     BASE=THETA(237)+ETA(5)
>     SLP=THETA(238)+ETA(6)
>     PSI=BASE+SLP*TIME
> ELSE IF (PSI_MODEL.EQ.4) THEN
>     BASE=THETA(239)+ETA(7)
>     SLP=THETA(240)+ETA(8)
>     PSI=BASE+SLP*TIME
> ELSE IF (PSI_MODEL.EQ.5) THEN
>     BASE=THETA(241)+ETA(9)
>     SLP=THETA(242)+ETA(10)
>     PSI=BASE+SLP*TIME
> END IF

to 

> IF (PSI_MODEL.EQ.1) THEN
>     BASE=THETA(233)+ETA(1)
>     SLP=THETA(234)+ETA(2)
>     PSI=BASE+SLP*TIME
> ENDIF 
> IF (PSI_MODEL.EQ.2) THEN
>     BASE=THETA(235)+ETA(3)
>     SLP=THETA(236)+ETA(4)
>     PSI=BASE+SLP*TIME
> ENDIF
> IF (PSI_MODEL.EQ.3) THEN
>     BASE=THETA(237)+ETA(5)
>     SLP=THETA(238)+ETA(6)
>     PSI=BASE+SLP*TIME
> ENDIF
> IF (PSI_MODEL.EQ.4) THEN
>     BASE=THETA(239)+ETA(7)
>     SLP=THETA(240)+ETA(8)
>     PSI=BASE+SLP*TIME
> ENDIF
> IF (PSI_MODEL.EQ.5) THEN
>     BASE=THETA(241)+ETA(9)
>     SLP=THETA(242)+ETA(10)
>     PSI=BASE+SLP*TIME
> ENDIF
rikardn commented 4 years ago

Thanks for reporting this. It is indeed a bug.