as.nm: Create and Manipulate nm Objects

In metrumrg: Pharmacometric Tools for Data Preparation, Modeling, Simulation, and Reporting

Description

Objects of class nm are intended to support analysis using the software NONMEM ((c), Icon Development Solutions). nm gives a zero-row data.frame with suitable columns and column classes (essentially, a template for dataset construction). as.nm and as.nm.data.frame construct an nm object from an existing object. The read and write methods are wrappers for ‘.csv’ equivalents. read.nm reconstitutes classes for flags, DATETIME, and C.

Usage

nm()
## S3 method for class 'data.frame'
as.nm(x, ...)
read.nm(x,na.strings='.',as.is=TRUE,key=c('SUBJ','TIME','SEQ'),flags=character(0),...)
write.nm(x, file, na = '.', row.names = FALSE, quote = FALSE, ...)

Arguments

x	data.frame or nm
na.strings	passed to read.csv
as.is	passed to read.csv
key	passed to as.keyed
flags	character vector naming colums to convert using as.flag
...	extra arguments, ignored or passed to write.csv
file	passed to write.csv
na	passed to write.csv
row.names	passed to write.csv
quote	passed to write.csv

Details

as.nm.data.frame is the principal method that creates an nm classification. It alone enforces all qualities of class nm. read.nm is the only other function that creates an nm classification; use with caution, as it does not enforce all qualities.

Just before as.nm.data.frame returns, it calls as.nm on each of its columns: a non-operation (as.nm.default) unless the user supplies specialized classes and methods. Column-specific methods may make use of the passed argument data, which contains the data set itself. Any enforced or conditional outputs (see column summary) have the column name as the first member of the class vector.

• SUBJ must be present and defined, even for commented records. ID is (re)calculated as as.numeric(factor(SUBJ)).

• C (class: comment) will be created if not present.

• NA C will be imputed FALSE.

• Every active (non-commented) source record should define exactly one of HOUR or DATETIME.

• HOUR is taken to represent relative accumulation of hours from arbitrary origin.

• DATETIME is understood as seconds, coercible to mDateTime.

• TIME is calculated from either HOUR or DATETIME.

• Definition (or not) of HOUR vs. DATETIME should be constant within subject (for active records).

• SEQ (class flag) will be created if not present.

• nm will be keyed on SUBJ, TIME, and SEQ. SEQ determines sort order for rows with matching TIME.

• Result will be sorted.

• TIME will be relativized to earliest extant value, incl. those in comments.

• TAFD (time after first non-commented dose), TAD (time since most recent non-commented dose), and LDOS (amount of most recent non-commented dose) will be calculated if AMT is present. TAD and LDOS are very literal: they return NA if no dose has been given yet, and will "remember" the time and amount of the most recent dose whether or not it is still conceptually relevant, given study design. If two doses are given at the same time, say in different compartments, LDOS reflects the first AMT with respect to sort order.

• TAD will consider ADDL and II if present.

• NA flags will be imputed as zero.

• MDV (missing dependent value) will be calculated if DV is present, preserving non-NA MDV, if present.

• resulting column order will lead with C followed by key columns.

Column summary:

• required inputs: SUBJ; HOUR or DATETIME

• optional inputs: AMT, ADDL, II, DV

• enforced outputs: SUBJ, ID, C, TIME, SEQ

• conditional outputs: TAFD, TAD, LDOS, MDV

Value

write.nm is used for side effects. Others return an object with class c('nm','keyed','data.frame').

Note

Assembly chains, such as nm() + dose + samp | demo, are no longer supported. Nor are as.moot and as.rigged.

Author(s)

Tim Bergsma

References

http://metrumrg.googlecode.com

See Also

• summary.nm

• Ops.keyed

Examples

## Not run: metrumrgURL('example/project/script/assemble.pdf')

dose <- data.frame( 
	SUBJ = rep(letters[1:3], each = 2), 
	HOUR = rep(c(0,20),3), 
	AMT = rep(c(40,60,80), each = 2) 
) 
dose <- as.keyed(dose,key=c('SUBJ','HOUR'))
samp <- data.frame( 
	SUBJ = rep(letters[1:3], each = 4), 
	HOUR = rep(c(0,10,20,30),3), 
	DV = signif(rnorm(12),2) + 2 
) 
samp <- as.keyed(samp,key=c('SUBJ','HOUR'))
demo <- data.frame( 
	SUBJ = letters[2:5], 
	RACE = c('asian','white','black','other'), 
	SEX = c('female','male','female','male'), 
	WT = c(75, 70, 73, 68) 
)
demo <- as.keyed(demo,key=c('SUBJ'))
meds <- as.keyed(
	data.frame(
		SUBJ=c('a','c'),
		HOUR=c(0,15),
		STOP=c(10,25),
		C3A4=as.flag(c(1,1))
	),
	key=c('SUBJ','HOUR')
)

nm()
#nm() + dose
as.nm(dose)
as.nm(dose + samp)
as.nm(dose + samp | demo) #as.nm executes once

meds
long <- deranged(meds,start='HOUR',stop='STOP')
long$EVID <- 2
as.nm( dose + samp + long)
data <- as.nm( aug(dose,EVID=1, SEQ=1) + aug(samp,EVID=0, SEQ=0) | demo)
summary(data,by=c('EVID','SEQ'))
## Not run: index(data)

metrumrg documentation built on May 2, 2019, 5:55 p.m.