LongToWide: LongToWide: Convert nested long structures to wide...

In google/glmmplus: Multiple Imputation and Model Selection

Description

In longitudinal or other multiple response studies, data presented in a long format will often feature dependence between rows. While this is the preferred format for lme4, such a format would hide important information from multiple imputation models and make the MAR assumption less plausible. Hense, the suggestion is to impute data in a wide format, where rows are again independent, and then return the mids object to a long format for use with FitModel, ForwardSelect, or BackwardEliminate.

Usage

WideToLong(data, id.name, response.base, time.varying.bases = c(""), sep = ".")

Arguments

data	A data frame or mids object in "long" format owing to multiple measurements within the same subject.
id.name	The subject id, a character string.
period.name

The repeated measurement (within subject) identifier. In a longitudinal study, this will be time.

time.varying.vars	A character vector of variable names that take multiple values per subject (in different rows)
sep	The character delimiter by which to separate the variable name base from the period identifier.

References

Stef van Buuren, Karin Groothuis-Oudshoorn (2011). mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software, 45(3), 1-67. URL http://www.jstatsoft.org/v45/i03/.

See Also

WideToLong

Examples

# Example of the long-to-wide, impute, wide-to-long strategy
library(glmmplus)
data(nls.97)
nls.97[1:10, 1:4]

nls.wide <- LongToWide(nls.97, id.name = "PUBID.1997", period.name = "age",
                       time.varying.vars = c("math.cs"))

nls.wide[1:2, c(1:2, 20:29)]
mids <- ImputeData(nls.wide, m = 5, maxit = 15, droplist = c("PUBID.1997"))
mids.long <- WideToLong(mids, "PUBID.1997", "math.cs")

google/glmmplus documentation built on May 17, 2019, 7:47 a.m.