build a column of data that maps baseline values to every row. baseline values starting points or reference points for each subject, e.g. a value of a measurement at time zero. The method will produce a mean value if multiple entries match the subjectID and visitID conditions.
fillBaselineColumn(
mxdfin,
columnName,
subjectID,
visitID,
baselineVisitValue,
baselineExt = "_BL",
deltaExt = "_delta",
fast = TRUE,
verbose = TRUE
)Input data frame with repeated measurements and a grouped time variable
string or vector of strings defining valid columns in the data frame to process.
the unique subject id column name
names of the column that defines the variable defining
baseline-ness. e.g. isBaseline
the value defining baseline e.g. TRUE
string appended to column name defining the baseline variable
string appended to column name defining the change variable
boolean; if TRUE, uses a faster, column-wise approach (requires data.table-like operations). Will only return subjects with baseline values. If there are several baseline entries, these will be averaged (for numeric) or the mode will be taken (for non-numeric). Only works robustly with atomic data types (numeric, character, factor).
boolean to print progress messages
data frame with new columns (or a list containing it if deltaExt is NA)
# Generate dummy data for an example
generateSubtyperData <- function(n_subjects = 10, n_visits = 5) {
set.seed(123)
subjects <- paste0("Sub_", 1:n_subjects)
df <- data.frame(
commonID = rep(subjects, each = n_visits),
yearsbl = rep(0:(n_visits - 1), n_subjects), # Assuming 0 is baseline
RandomBasisProjection01 = rnorm(n_subjects * n_visits, mean = 10, sd = 2),
CategoricalVar = sample(c("A", "B", "C"), n_subjects * n_visits, replace = TRUE),
stringsAsFactors = FALSE
)
# Introduce some missing baseline data for testing 'fast' mode filtering
df$RandomBasisProjection01[sample(1:nrow(df), 5)] <- NA
# Introduce multiple baselines for some subjects
extra_baseline_subject <- df[df$commonID == subjects[1] & df$yearsbl == 0,]
extra_baseline_subject$yearsbl <- 0.001 # slightly different baseline visit
df <- rbind(df, extra_baseline_subject)
return(df)
}
# Example 1: Single numeric column with fast mode
mydf <- generateSubtyperData(10)
mydog <- fillBaselineColumn(
mxdfin = mydf,
columnName = "RandomBasisProjection01",
subjectID = 'commonID',
visitID = 'yearsbl',
baselineVisitValue = 0,
fast = TRUE,
verbose = TRUE
)[[1]]
#> Using fast path...
#> Filtering mxdfin to subjects with baseline data.
#> Replicating baseline rows...
#> Filling columns...
#> Fill done.
# Check a subject with multiple baselines (Sub_1, has duplicate for yearsbl=0)
# Note: The example data generation for multiple baselines might need manual adjustment
# to realistically simulate it for 'aggregate' to take effect.
# For instance, if Sub_1 has two 'yearsbl' == 0 entries with different
# RandomBasisProjection01 values.
# Example 2: Multiple columns (numeric and categorical) with fast mode
mydf2 <- generateSubtyperData(10)
mydog2 <- fillBaselineColumn(
mxdfin = mydf2,
columnName = c("RandomBasisProjection01", "CategoricalVar"), # vars2bl from your example
subjectID = 'commonID',
visitID = 'yearsbl',
baselineVisitValue = 0,
fast = TRUE,
verbose = FALSE
)[[1]]
head(mydog2)
#> commonID yearsbl RandomBasisProjection01 CategoricalVar
#> 1 Sub_1 0.000 8.879049 A
#> 2 Sub_1 1.000 9.539645 B
#> 3 Sub_1 2.000 13.117417 A
#> 4 Sub_1 3.000 10.141017 B
#> 5 Sub_1 4.000 10.258575 A
#> 51 Sub_1 0.001 8.879049 A
#> RandomBasisProjection01_BL CategoricalVar_BL RandomBasisProjection01_delta
#> 1 8.879049 A 0.0000000
#> 2 8.879049 A 0.6605963
#> 3 8.879049 A 4.2383679
#> 4 8.879049 A 1.2619681
#> 5 8.879049 A 1.3795268
#> 51 8.879049 A 0.0000000
#> CategoricalVar_delta
#> 1 NA
#> 2 NA
#> 3 NA
#> 4 NA
#> 5 NA
#> 51 NA
# Example 3: Fast mode, specific scenario if a subject entirely lacks baseline
# mydf3 <- generateSubtyperData(5)
# mydf3 <- mydf3[!(mydf3$commonID == "Sub_1" & mydf3$yearsbl == 0), ] # Remove Sub_1's baseline
# mydog3 <- fillBaselineColumn(
# mxdfin = mydf3,
# columnName = "RandomBasisProjection01",
# subjectID = 'commonID',
# visitID = 'yearsbl',
# baselineVisitValue = 0,
# fast = TRUE,
# verbose = TRUE
# )[[1]]
# print(unique(mydog3$commonID)) # Sub_1 should be absent
# Example 4: Slow mode (default)
mydf_slow <- generateSubtyperData(5)
mydog_slow <- fillBaselineColumn(
mxdfin = mydf_slow,
columnName = "RandomBasisProjection01",
subjectID = 'commonID',
visitID = 'yearsbl',
baselineVisitValue = 0,
fast = FALSE, # Explicitly use the slow path
verbose = TRUE
)[[1]]
#> Using slow path...
#> 0%. 100%.
head(mydog_slow)
#> commonID yearsbl RandomBasisProjection01 CategoricalVar
#> 1 Sub_1 0 8.879049 A
#> 2 Sub_1 1 9.539645 C
#> 3 Sub_1 2 13.117417 A
#> 4 Sub_1 3 10.141017 C
#> 5 Sub_1 4 10.258575 B
#> 6 Sub_2 0 13.430130 A
#> RandomBasisProjection01_BL RandomBasisProjection01_delta
#> 1 8.879049 0.0000000
#> 2 8.879049 0.6605963
#> 3 8.879049 4.2383679
#> 4 8.879049 1.2619681
#> 5 8.879049 1.3795268
#> 6 13.430130 0.0000000