To start, load the package.
Logical Comparisons
This section covers basic logical comparisons and shows how they
might be done in base R versus using the
extraoperators package. Many of these are quite simple, but
are defined so that later operators are possible.
First let’s define our “data” as some numbers stored in
sample_numbers.
sample_numbers <- c(9, 1, 5, 3, 4, 10, 99)Now we can do a series of simple logical comparisons which return a
logical vector of TRUE or FALSE.
## base R: greater than 3?
sample_numbers > 3
#> [1] TRUE FALSE TRUE FALSE TRUE TRUE TRUE
## base R: greater than or equal to 3?
sample_numbers >= 3
#> [1] TRUE FALSE TRUE TRUE TRUE TRUE TRUE
## base R: less than 3?
sample_numbers < 3
#> [1] FALSE TRUE FALSE FALSE FALSE FALSE FALSE
## base R: less than or equal to 3?
sample_numbers <= 3
#> [1] FALSE TRUE FALSE TRUE FALSE FALSE FALSEUnfortunately, we cannot use < or >
in custom operators, so we use the substitutions: g = >
and l = < and e = =. So that
ge = <= etc.
## extraoperators: greater than 3?
sample_numbers %g% 3
#> [1] TRUE FALSE TRUE FALSE TRUE TRUE TRUE
## extraoperators: greater than or equal to 3?
sample_numbers %ge% 3
#> [1] TRUE FALSE TRUE TRUE TRUE TRUE TRUE
## extraoperators: less than 3?
sample_numbers %l% 3
#> [1] FALSE TRUE FALSE FALSE FALSE FALSE FALSE
## extraoperators: less than or equal to 3?
sample_numbers %le% 3
#> [1] FALSE TRUE FALSE TRUE FALSE FALSE FALSESo far there is no real gain in using extraoperators but
this changes for more complex operations. What if we want to know if our
values fall within some range? This is a fairly common task, such as
saying that valid ages must be between 0 and 100 years.
## base R: greater than 3 and less than 10?
sample_numbers > 3 & sample_numbers < 10
#> [1] TRUE FALSE TRUE FALSE TRUE FALSE FALSE
## base R: greater than or equal to 3 and less than 10?
sample_numbers >= 3 & sample_numbers < 10
#> [1] TRUE FALSE TRUE TRUE TRUE FALSE FALSE
## base R: greater than 3 and less than or equal to 10?
sample_numbers > 3 & sample_numbers <= 10
#> [1] TRUE FALSE TRUE FALSE TRUE TRUE FALSE
## base R: greater than or equal to 3 and less than or equal to 10?
sample_numbers >= 3 & sample_numbers <= 10
#> [1] TRUE FALSE TRUE TRUE TRUE TRUE FALSEBase R accomplishes this through chaining of operations.
extraoperators has built in range operators.
## extraoperators: greater than 3 and less than 10?
sample_numbers %gl% c(3, 10)
#> [1] TRUE FALSE TRUE FALSE TRUE FALSE FALSE
## extraoperators: greater than or equal to 3 and less than 10?
sample_numbers %gel% c(3, 10)
#> [1] TRUE FALSE TRUE TRUE TRUE FALSE FALSE
## extraoperators: greater than 3 and less than or equal to 10?
sample_numbers %gle% c(3, 10)
#> [1] TRUE FALSE TRUE FALSE TRUE TRUE FALSE
## extraoperators: greater than or equal to 3 and less than or equal to 10?
sample_numbers %gele% c(3, 10)
#> [1] TRUE FALSE TRUE TRUE TRUE TRUE FALSEFinally, extraoperators includes a not in
operator, %!in%.
## base R: not in 3 or 10
!sample_numbers %in% c(3, 10)
#> [1] TRUE TRUE TRUE FALSE TRUE FALSE TRUE
## extraoperators: not in 3 or 10
sample_numbers %!in% c(3, 10)
#> [1] TRUE TRUE TRUE FALSE TRUE FALSE TRUEThe next sections show a few examples of these operators augmented by
prefixes: ?, s and a.
Indices (Which Values?)
Sometimes we want to use a logical comparison and identify indices,
such as to use in a loop. extraoperators does this by
prefixing operators with ? for “which”.
## base R: what are the indices that match 3 and 10?
which(sample_numbers %in% c(3, 10))
#> [1] 4 6
## extraoperators: what are the indices that match 3 and 10?
sample_numbers %?in% c(3, 10)
#> [1] 4 6
## base R: what are the indices for numbers between 3 and 10?
which(sample_numbers > 3 & sample_numbers < 10)
#> [1] 1 3 5
## extraoperators: what are the indices for numbers between 3 and 10?
sample_numbers %?gl% c(3, 10)
#> [1] 1 3 5This can be readily incorporated in other code for further processing.
Subsetting
Another fairly common task is selecting only certain observations.
For example, we might want to calculate the average of numbers within a
plausible range (e.g., excluding outliers). In
extraoperators subsetting is done by adding an
s prefix.
## base R: subset to only numbers between 3 and 10
mean(subset(sample_numbers, sample_numbers > 3 & sample_numbers < 10))
#> [1] 6
## or equivalently
mean(sample_numbers[sample_numbers > 3 & sample_numbers < 10])
#> [1] 6
## extraoperators: subset to only numbers between 3 and 10
mean(sample_numbers %sgl% c(3, 10))
#> [1] 6Subsetting can be especially useful in quick exploratory analyses. Graphs are easily hard to read if there are extreme values. Subsetting makes it fast to “zoom in” on a specific range.
All (or None)
Finally, you might have some quality controls in place for data. For
example asserting that all ages are between 0 and 100. In
extraoperators this is done by adding the prefix
a.
## base R: are all numbers between 0 and 10?
all(sample_numbers > 0 & sample_numbers < 10)
#> [1] FALSE
## extraoperators: are all numbers between 0 and 10?
sample_numbers %agl% c(0, 10)
#> [1] FALSE
## extraoperators: are all numbers between 0 and 100?
sample_numbers %agl% c(0, 100)
#> [1] TRUEIf you want to know the opposite, are no numbers between 0 and 100, we can negate the whole operation.
## extraoperators: are NO numbers between 0 and 100?
!sample_numbers %agl% c(0, 100)
#> [1] FALSE
## extraoperators: are NO numbers between 55 and 60?
!sample_numbers %agl% c(55, 60)
#> [1] TRUEThere are also expanded all, subset, and which operators for equals and not equals.
Chaining
In language, it is fairly natural to make a statement like this: “In
my study, age should be between 18 to 65 and not be missing.” In
R, the usual implementation of this is more equivalent to:
“In my study, age should be greater than 18 and age should be less than
65 and age should not be missing.” extraoperators tries to
facilitate something closer to the cleaner original statement using the
chaining operator, %c%. The chaining operator chains a set
of operations on the right hand side with the argument on its left hand
side passed to each. To accomplish this, the right hand side must be
quoted.
age <- c(19, 30, 90, 50, NA, 45)
age %c% "(> 18 & < 65) & !is.na"
#> [1] TRUE TRUE FALSE TRUE FALSE TRUEBecause the right hand side of the chaining operator is a character
string that is parsed, it is possible to do some special things in it.
is.na, !is.na, is.nan, and
!is.nan are special characters that do not require any
further value to work correctly. As shown, parentheses also work, which
allows fine grained control over exactly what is intended. For example
if we expect only adults are in the study, but those who refused to
report their age were coded -9 and people who failed to complete the
questionnaire at all are missing.
age <- c(19, 30, 90, 50, NA, 16, -9)
age %c% "(> 18 | == -9) & !is.na"
#> [1] TRUE TRUE TRUE TRUE FALSE FALSE TRUEAs with all operators, there are prefixes for all, subset, and which.
Interval Notation Operator
In math, interval notation often is used. For example, we might
write: \(x \in (1, 5) \cup [6,
\infty)\) to indicate that x is between the intervals 1
to 5 (not including 1 or 5) or between 6 and positive infinity,
including 6 but not positive infinity. The interval notation operator,
%e% let’s you use fairly similar language in
R. “|” is the union operator and “&” is the intersect
operator. Variables are allowed but no functions as these cannot be
parsed.
Regular Expressions
Sometimes you want to pattern match. For example, you might want to
find all variable names that match a certain pattern. The
%grepl% operator can help here, built offR'sgrepl`
function.
## sample dataset
data <- data.frame(
ID = c(1, 2, 3),
cesd_1 = c(4, 5, 6),
cesd_2 = c(7, 8, 9),
cesd_total = c(11, 13, 15)
)
## find all variables that start with "cesd"
names(data)[grepl("^cesd", names(data))]
#> [1] "cesd_1" "cesd_2" "cesd_total"
## or equivalently using grep() with right options
grep("^cesd", names(data), value = TRUE)
#> [1] "cesd_1" "cesd_2" "cesd_total"
## here is the operator version
names(data) %sgrepl% "^cesd"
#> [1] "cesd_1" "cesd_2" "cesd_total"
## the operator opens up all standard variations
names(data) %?grepl% "^cesd" ## indices
#> [1] 2 3 4
names(data) %s!grepl% "^cesd" ## subset names not in pattern
#> [1] "ID"
names(data) %agrepl% "^cesd" ## do all match the pattern?
#> [1] FALSEAn example use case in practice might be to find all variables that are items in a scale and then calculate the scale score.
set.seed(123) # Set seed for reproducibility
data <- data.frame(matrix(sample(0:4, 100, replace = TRUE), ncol = 10))
names(data) <- paste0("cesd_", 1:10)
data <- cbind(ID = 1:10, data)
## find all variables that start with "cesd" and end with a number
## use these to sum all the items for the scale score
data$cesd_total <- rowSums(data[, names(data) %sgrepl% "^cesd.*[0-9]$"])
summary(data$cesd_total)
#> Min. 1st Qu. Median Mean 3rd Qu. Max.
#> 12.0 18.0 19.0 19.3 20.5 25.0Another example is you load a larger dataset from someone and want to check whether there are any variable names including spaces. You want to check that any space, in regular expression captured by “\s”, is not contained in any variable name. That is all are not in space.
This returns TRUE indicating no variable names have a
space in them.
Here is an example that fails. At least one variable name has a
space. We follow it up by finding out which variable(s). This
could be fed back to the data owners to change, if desired. Lastly,
although these are written as separate codes, when used interactively,
one might start with the test and if it fails change the a!
to s to find the variables. The intention being just to
slightly ease and speed up the process.
data <- data.frame(
a = 1:4,
`b c` = 5:8, check.names = FALSE
)
names(data) %a!grepl% "\\s" # this fails
#> [1] FALSE
# which variables have spaces?
names(data) %sgrepl% "\\s"
#> [1] "b c"