Change detection of eta type variables

[MikeKSmith]

In read_nm_tables.R function index_table the variable detection for type eta is
stringr::str_detect(.$col, 'ETA\d+|ET\d+') ~ 'eta',

Consider changing to
stringr::str_detect(.$col, '^ETA|^ET') ~ 'eta'

to allow detection of ETA parameters by name e.g.
ETA_CL, ETA_V etc.

[bguiastr]

Thanks for the suggestion.
My initial thoughts were to be rather strict to avoid false positive such as THETA. In that sense I think it is good that you added the regex '^', I just make one suggestion, could you please change '^ET' to '^ET\d+'?

This would however bring a problem to the eta_distrib() and eta_qq() plots due to use of a post-processing function to make the labels look nicer and in the right order. We will need to have an if statement in these functions checking that if not all 'eta' type variable match the 'ETA\d+|ET\d+' the post-processing function should be removed.

[jeslo328]

Hi Ben

I used this package - and it was easy to be maneuver as well adjustable. (fantastic job Ben)
So how would one change the "eta" to its respective names.
Will i be able to compile all the diagnostic plot in 1 as in old xpose (where we would be able to do with basic.gof(xpdb))

Thank you
Jesmin

[billdenney]

I just had the same issue as @MikeKSmith with my ETAs named: ETA(KA), ETA(CL), ETA(V1), ETA(Q2), ETA(V2), ETA(Q3), ETA(V3)

[bguiastr]

@jeslo328 I guess an easy way to rename your panels would be through the labeller function:

xpdb_ex_pk %>% 
eta_distrib(xpdb = xpdb_ex_pk,
            labeller = labeller(variable = c('ETA(1)' = 'ETA CL',
                                             'ETA(2)' = 'ETA Vc',
                                             'ETA(3)' = 'ETA Ka')))

[bguiastr]

@MikeKSmith and @billdenney I had not realized that so many modelers were using specific ETA names. I can see the value for this in the plot and that it need to be implemented in the next version of xpose.

FYI the current code was developed to make use of $TABLE ETAS(1:LAST) in NONMEM which I find useful as you never forget an ETA value. I would also like for xpose to be able to make use of the phi file if no $TABLE are available... with a .source

[billdenney]

@guiastrennec, I have moved to almost always using THETA and ETA naming so that I don't have to remember what the numbers of the parameters are. And, I made a suggestion to Bob Bauer a few weeks ago that will make them even more usable and useful; in the next version of NONMEM, you will be able to specify something like the following (syntax may be slightly different)

$THETA
CL=(0, 1, 5)

which will expand to

$THETA
(0, 1, 5)
$ABBR REPLACE THETA(CL)=THETA(1)

The same format will be available for $OMEGA and $SIGMA.

After that is available, I doubt that I will ever use numbers again.

[bguiastr]

Thanks for letting me know about this, it is very interesting!