Mark units that have a large x
mark_the_largest.RdA function that marks the units that have the largest x-value in a data set. Can mark the largest ones so that they cover a specified percentage of total x (method 1), can mark values that are greater than a specified threshold (method 2), can mark the n largest values (method 3) and can mark the largest so a specified percent of all units are marked (method 4).
Usage
mark_the_largest(
data,
idVar,
strataVar = NULL,
xVar,
yVar = NULL,
method = 1,
par_method1 = NULL,
par_method2 = NULL,
par_method3 = NULL,
par_method4 = NULL,
max_n_method1and2 = NULL,
min_x_method3and4 = NULL
)Arguments
- data
Input data set of class data.frame
- idVar
Name of identification variable. Should not have NA's, and should not have duplicates
- strataVar
Optional. Name of stratification variable. Should not have NA's. If strataVar is given, the marking is performed within each stratum
- xVar
Name of x-variable. Should be numeric. +/-Inf is not allowed. NA's are allowed (would never be marked)
- yVar
Optional. Name of an extra sorting variable. Should be numeric. Only relevant for the sorting of equal x-values, in which case the x's are ranked according to decreasing y-value. NA's and +/-Inf are allowed. (NA is rated as less than -Inf)
- method
The methods to be used (default is 1). Can choose between method 1, 2, 3 and 4 (see ‘Details’), and can use multiple methods simultaneously. The methods are specified using a vector
- par_method1
Percentage for method 1 (default 25). Numeric value within the range [0, 100] (a single number or a vector with a length equal to the number of strata). If 0, no values are marked. If 100, all x > 0 are marked (assuming max_n_method1and2 is not used)
- par_method2
Threshold value for method 2. Numeric value >= 0 (a single number or a vector with a length equal to the number of strata). If 0, all x > 0 are marked (assuming max_n_method1and2 is not used)
- par_method3
Number for method 3 (default 5). Integer 0, 1, 2, ... (a single number or a vector with a length equal to the number of strata). If 0, no values are marked
- par_method4
Percentage for method 4 (default 5). Numeric value within the range [0, 100] (a single number or a vector with a length equal to the number of strata). If 0, no values are marked. If 100, all x != NA are marked (assuming min_x_method3and4 is not used)
- max_n_method1and2
Optional. Applies to method 1 and 2. Maximum number of markings that can be made. Integer 0, 1, 2, ...(a single number or a vector with a length equal to the number of strata)
- min_x_method3and4
Optional. Applies to method 3 and 4. Minimum threshold for x (only x > min_x_method3and4 can be marked). Numeric (a single number or a vector with a length equal to the number of strata)
Value
The output is a data frame. For each method used, a binary variable is created that shows which units are marked (the variables are named large1, large2, large3, and large4 for methods 1, 2, 3 and 4 respectively). Additionally, there is a variable that shows which units are marked for at least one method (the variable is named large).
Details
If strataVar is given, the marking is performed within each stratum. Parameters given as a single number then apply to
each stratum. If a parameter is given as a vector, the length should equal to the number of strata, and the order of
the elements should correspond to the order obtained when the input data is sorted by strataVar using order():
data[order(data[ , strataVar]) , ]
Method 1: Marks the largest x-values so that they (at least) cover a specified percentage of the total x-value (x < 0 are not included in the total x-value).
Only x > 0 can be marked with this method.
If strataVar is used, this is done per stratum, and the total x-value applies to the stratum.
If max_n_method1and2 is used, it is not guaranteed that the specified percentage will be achieved.
Method 2: Marks x-values that are greater than a specified threshold (x > threshold).
Only x > 0 can be marked with this method.
If max_n_method1and2 is used, it's not guaranteed that all x > threshold will be marked.
Method 3: Marks the n largest x-values.
x = NA will never be marked with this method.
If strataVar is used, it's the n largest per stratum that are marked.
If min_x_method3and4 is used, it's not guaranteed that the specified number will be marked.
Method 4: Marks the largest x-values so that p percent of all units are marked (units with x = NA are not counted in the percentage calculation).
x = NA will never be marked with this method.
If strataVar is used, the p percent largest units per stratum are marked.
If min_x_method3and4 is used, it's not guaranteed that the specified percentage will be achieved.
To ensure the same result regardless of the sorting of the input data set, the function sorts the input data set
first by idVar before further sorting and marking is done.
Examples
# Test dataset
set.seed(956)
testData <- data.frame(id = 1:30, x = rnorm(n = 30, mean = 1000, sd = 500),
strata = c(rep('A', 12), rep('B', 18)))
testData$x[sample(1:30, size = 2)] <- NA
# Example with method 1, 2 and 3
mark_the_largest(data = testData, idVar = 'id', xVar = 'x', method = c(1, 2, 3),
par_method1 = 15, par_method2 = 5000, par_method3 = 5)
#> id x par_method1 par_method2 par_method3 large1 large2 large3 large
#> 1 28 1755.82152 15 5000 5 1 0 1 1
#> 2 3 1719.23612 15 5000 5 1 0 1 1
#> 3 1 1688.47966 15 5000 5 1 0 1 1
#> 4 24 1553.10406 15 5000 5 0 0 1 1
#> 5 17 1395.47290 15 5000 5 0 0 1 1
#> 6 4 1348.71579 15 5000 5 0 0 0 0
#> 7 30 1268.94547 15 5000 5 0 0 0 0
#> 8 5 1266.01136 15 5000 5 0 0 0 0
#> 9 14 1174.49728 15 5000 5 0 0 0 0
#> 10 13 1150.16903 15 5000 5 0 0 0 0
#> 11 23 1137.08342 15 5000 5 0 0 0 0
#> 12 12 1081.14497 15 5000 5 0 0 0 0
#> 13 22 1038.01792 15 5000 5 0 0 0 0
#> 14 2 984.35659 15 5000 5 0 0 0 0
#> 15 16 959.67966 15 5000 5 0 0 0 0
#> 16 29 914.75150 15 5000 5 0 0 0 0
#> 17 19 788.83780 15 5000 5 0 0 0 0
#> 18 11 700.85778 15 5000 5 0 0 0 0
#> 19 25 610.86377 15 5000 5 0 0 0 0
#> 20 10 601.84912 15 5000 5 0 0 0 0
#> 21 20 534.26997 15 5000 5 0 0 0 0
#> 22 8 476.92625 15 5000 5 0 0 0 0
#> 23 9 451.53607 15 5000 5 0 0 0 0
#> 24 7 422.98241 15 5000 5 0 0 0 0
#> 25 6 305.84389 15 5000 5 0 0 0 0
#> 26 18 113.40098 15 5000 5 0 0 0 0
#> 27 21 -44.53893 15 5000 5 0 0 0 0
#> 28 26 -80.36282 15 5000 5 0 0 0 0
#> 29 15 NA 15 5000 5 0 0 0 0
#> 30 27 NA 15 5000 5 0 0 0 0
# Example with stratification
mark_the_largest(data = testData, idVar = 'id', strataVar = 'strata', xVar = 'x',
method = c(1, 4), par_method1 = c(20, 30), par_method4 = 25)
#> id x strata par_method1 par_method4 large1 large4 large
#> 1 3 1719.23612 A 20 25 1 1 1
#> 2 1 1688.47966 A 20 25 1 1 1
#> 3 4 1348.71579 A 20 25 0 1 1
#> 4 5 1266.01136 A 20 25 0 0 0
#> 5 12 1081.14497 A 20 25 0 0 0
#> 6 2 984.35659 A 20 25 0 0 0
#> 7 11 700.85778 A 20 25 0 0 0
#> 8 10 601.84912 A 20 25 0 0 0
#> 9 8 476.92625 A 20 25 0 0 0
#> 10 9 451.53607 A 20 25 0 0 0
#> 11 7 422.98241 A 20 25 0 0 0
#> 12 6 305.84389 A 20 25 0 0 0
#> 13 28 1755.82152 B 30 25 1 1 1
#> 14 24 1553.10406 B 30 25 1 1 1
#> 15 17 1395.47290 B 30 25 1 1 1
#> 16 30 1268.94547 B 30 25 0 1 1
#> 17 14 1174.49728 B 30 25 0 0 0
#> 18 13 1150.16903 B 30 25 0 0 0
#> 19 23 1137.08342 B 30 25 0 0 0
#> 20 22 1038.01792 B 30 25 0 0 0
#> 21 16 959.67966 B 30 25 0 0 0
#> 22 29 914.75150 B 30 25 0 0 0
#> 23 19 788.83780 B 30 25 0 0 0
#> 24 25 610.86377 B 30 25 0 0 0
#> 25 20 534.26997 B 30 25 0 0 0
#> 26 18 113.40098 B 30 25 0 0 0
#> 27 21 -44.53893 B 30 25 0 0 0
#> 28 26 -80.36282 B 30 25 0 0 0
#> 29 15 NA B 30 25 0 0 0
#> 30 27 NA B 30 25 0 0 0