Extract columns that vary within values of other columns in a data.frame

If you want to look at the variability of a number of columns and you want to disregard those that are constant. Like for findCovs, by can be of arbitrary length.

findVars(data, by = NULL, return.data = TRUE, as.fun = NULL, cols.id)

Arguments

data

data.frame in which to look for covariates

by

optional covariates will be searched for in combinations of values in these columns. Often by will be either empty or ID. But it can also be both say c("ID","DRUG") or c("ID","TRT").

return.data

Return the data set with columns that vary? If FALSE, only the column names will be returned as a character vector. Default is TRUE.

as.fun

The default is to return a data.table if data is a data.table and return a data.frame in all other cases. Pass a function in as.fun to convert to something else. If data is not a data.table, the default can be configured using `NMdataConf()`.

cols.id

Deprecated. Use by instead.

Value

a data set with as many rows as in data. If return.data is FALSE, a character vector with column names.

Details

Use this to exclude columns that are constant within by. If by=ID, this could be to get only time-varying covariates.

See also

Other DataCreate: NMorderColumns(), NMstamp(), NMwriteData(), addTAPD(), findCovs(), flagsAssign(), flagsCount(), mergeCheck(), tmpcol()

Examples

dt1 <- data.frame(ID=c(1,1,2,2),
                  OCC=c(1,2,1,2),
               ## ID level
                  eta1=c(1,1,3,3),
               ## occasion level
                  eta2=c(1,3,1,5),
               ## not used
                  eta3=0
               )
## model level
findCovs(dt1)
#>   eta3
#> 1    0
## ID level
findCovs(dt1,"ID")
#>   ID eta1 eta3
#> 1  1    1    0
#> 2  2    3    0
## acual ID level
findVars(findCovs(dt1,"ID"))
#>   ID eta1
#> 1  1    1
#> 2  2    3
## occasion level
findCovs(findVars(dt1,"ID"),c("ID","OCC"))
#>   ID OCC eta2
#> 1  1   1    1
#> 2  1   2    3
#> 3  2   1    1
#> 4  2   2    5