Descriptive aliases for ETAs

[kylebaron]

For example:

$OMEGA 
name="PK"
labels=c(CL,VC,KA)
1 2 3

$OMEGA
name="PD"
labels=c(IC50,KIN)
4 5 

$MAIN
double CLi = TVCL*exp(ETA_PK_CL);
double IC50i = TVIC50*exp(ETA_PD_IC50);

[dpastoor]

i'd throw out a minor syntatic proposal of instead of name be like prefix, with a default value of ETA_ such that

$OMEGA 
labels=c(CL,VC,KA)
1 2 3

$MAIN
double CLi = TVCL*exp(ETA_CL);
double IC50i = TVIC50*exp(ETA_IC50);

however could be overridden

$OMEGA 
prefix="ETA_PK"
labels=c(CL,VC,KA)
1 2 3

$MAIN
double CLi = TVCL*exp(ETA_PK_CL);

or completely ignored if someone really wanted

$OMEGA 
prefix=""
labels=c(ECL,EVC,EKA)
1 2 3



$MAIN
double CLi = TVCL*exp(ECL);

[kylebaron]

Nice ... will do it. Thanks!

[kylebaron]

Screen Shot 2016-03-17 at 11.28.41 PM.pdf

[kylebaron]

This is how it's looking; and passing all the unit tests I already had in place. Trying to see where the other gotchas are going to be here. @vjd

• Since we took the time to set up labels for each omega element, we can now more-easily "capture" the value
• It's still okay not to include labels for one of the random effect blocks
• Included the prefix piece but as @dpastoor noted people are free to omit that and code it directly into the label
• Error if you do provide labels and the number of unique labels provided doesn't match the number of rows for the matrix
• Error if one of the labels is the same as a parameter name or a compartment name (that would have resulted in a complaint from the compiler anyway)
• Same setup for $SIGMA
• oprefix, olabels, sprefix, slabels are now arguments for $NMXML
• Probably need an error if someone uses punctuation other than _

$PARAM TVCL = 1, TVVC=30, TVKA=1.1

$CMT GUT CENT

$ADVAN2

$MAIN
pred_CL = TVCL*exp(ECL);
pred_VC = TVVC*exp(EVC);
pred_KA = TVKA*exp(EKA);

$OMEGA
prefix="E", labels = s(CL,VC,KA)
0.1 0.2 0.3

$CAPTURE ECL EKA TVCL pred_CL

Also note that options are evaluated so you could do this (if you wanted):

$OMEGA
labels=paste0("ETA_", s(CL,VC,KA))
1 2 3
$MAIN
pred_CL = TVCL*exp(ETA_CL);
// ....

I don't see it usually working like that, but you could do it.

[vjd]

This is a helpful suggestion @dpastoor and thanks for implementing this right away @kylebmetrum

Kyle - you presumably take the same concept into $CMT that will allow extraction of the compartment numbers easier when referring _F(n), _ALG(n) etc. I understand that right now we have to make a call to check the compartment number (which is sequential by default), but this can be really useful for systems biology kind of models.

[kylebaron]

I didn't update #13 yet, but here's what I did:

$CMT DEPOT CENT PERIPH

$MAIN

F_DEPOT = 0.5;

D_CENT = 2.3;

ETC .... so you will get alias for F_CMT, D_CMT, R_CMT and ALAG_CMT. I'm pretty sure this will end use of compartment numbers in the specification file.

From the data side, you can do this:

data <- expand.ev(amt=1000, cmt=cmtn(mod,"DEPOT"))

Or (not documented)

mod %>% ev(amt=1000, cmt="DEPOT") %>% mrgsim %>% ...

So if cmt is character in the data set, we look up the compartment number for you and replace with that integer.