block-format: Block Format

In benjaminrich/PCSmisc: Useful functions for longitudinal data set construction and manipulation.

Block Format

Description

Block format is the typical ‘long’ format for longitudinal data, whereby the data for an individual is represented by a block of rows in a data set, and an ID column (or in some cases multiple columns) serves to delineate the blocks. It is also called ‘stacked’ data format.

For the purposes of the PCSmisc library, a block format is defined through an ID, which must be a factor and must possess the property of being sorted: i.e. for a variable id to be a valid ID, the following expression must evaluate to TRUE: is.factor(id) && !is.unsorted(id).

This strict requirement is intended to reduce the likelihood of committing errors when using functions in PCSmisc that operate on data in block format. In case the column(s) that define the blocks in a particular data set do not conform to these requirements, a valid ID can easily be constructed using asID.

The size of a block is the number of times the corresponding ID appears in the data. A block can also have size zero. This occurs when there are levels of the ID that do not appear in the data at all. The size of each block can most easily be obtained using blk.count.

PCSmisc contains a number of functions that operate on data in block format. These functions are identifiable by the prefix blk. and are generally wrappers of tapply. Like tapply, these functions do not operate on data.frame's directly but rather on their columns (i.e. on individual vectors of the same length). The purpose of these functions is to provide abstractions for common data operations, resulting in code that is easier to understand and reducing the likelihood of errors. Some of these functions are quite general, others are more specific to the creation of population PK type datasets, as required e.g. for analysis with NONMEM. See examples below.

Author(s)

Benjamin Rich <[e-mail]>

See Also

asID blk.count blk.repeatValue tapply

Examples

require(nlme)
data(Phenobarb)
dat <- Phenobarb[1:56,]  # First 4 subjects

attach(dat)

is.sorted(Subject)          # FALSE
is.sorted(asID(Subject))    # TRUE

data.frame(id=asID(Subject),
           time=time,
           dose=dose,
           first.dose=blk.repeatValue(dose, asID(Subject), ind=(!is.na(dose))),
           conc=conc,
           count.conc=blk.count(asID(Subject), ind=!is.na(conc)),
           last.dose=blk.locf(dose, asID(Subject)),
           time.after.dose=blk.tad(time, asID(Subject), !is.na(dose)))

detach(dat)

### The size of each block (example)
id <- factor(c("A", "A", "C", "C", "C"), levels=c("A", "B", "C"))
blk.count(id, NULL)  # Note that "B" is a valid ID.  The corresponding block has size zero.

benjaminrich/PCSmisc documentation built on Feb. 11, 2024, 9:25 p.m.