Package 'sift'

Join tables based on overlapping intervals.

Description

User-friendly interface that synthesizes power of dplyr::left_join and findInterval.

Usage

break_join(x, y, brk = character(), by = NULL, ...)

Arguments

x	A data frame.
y	Data frame containing desired reference information.
brk	Name of column in x and y to join by via interval overlapping. Must be coercible to numeric.
by	Joining variables, if needed. See mutate-joins.
...	additional arguments automatically directed to findInterval and dplyr::left_join. No partial matching.

Value

An object of the same type as x.

• All x rows will be returned.

• All columns between x and y are returned.

• Rows in y are matched with x based on overlapping values of brk (e.g. findInterval(x$brk, y$brk, ...)).

Examples

# joining USA + UK leaders with population time-series
break_join(us_uk_pop, us_uk_leaders, brk = c("date" = "start"))

# simple dataset
set.seed(1)
a <- data.frame(p = c(rep("A", 10), rep("B", 10)), q = runif(20, 0, 10))
b <- data.frame(p = c("A", "A", "B", "B"), q = c(3, 5, 6, 9), r = c("a1", "a2", "b1", "b2"))

break_join(a, b, brk = "q") # p identified as common variable automatically
break_join(a, b, brk = "q", by = "p") # same result
break_join(a, b, brk = "q", all.inside = TRUE) # note missing values have been filled

# joining toll prices with vehicle time-series

library(mopac)
library(dplyr, warn.conflicts = FALSE)
library(hms)

express %>%
  mutate(time_hms = as_hms(time)) %>%
  break_join(rates, brk = c("time_hms" = "time"))

---

Simulated records of radio station communications.

Description

Dataset intended to demonstrate usage of sift::conjecture.

Usage

comms

Format

An object of class tbl_df (inherits from tbl, data.frame) with 50000 rows and 4 columns.

---

Specialized "long to wide" reshaping

Description

On the surface, conjecture() appears similar to tidyr::pivot_wider(), but uses different logic tailored to a specific type of dataset:

• column corresponding to names_from contains only 2 levels

• there is no determinate combination of elements to fill 2 columns per row.

See vignette("conjecture") for more details.

Usage

conjecture(data, sort_by, names_from, names_first)

Arguments

data	A data frame to reshape.
sort_by	Column name, as symbol. Plays a similar role as values_from in pivot_wider(), but also serves as sorting dimension for underlying conjecture algorithm.
names_from	Column name, as symbol. Used to differentiate anterior/posterior observations. Column must only contain 2 levels (missing values not allowed).
names_first	level in variable specified by names_from indicating anterior observation.

Details

conjecture() uses the following routine to match elements:

1. Values in sort_by are separated into two vectors: anterior and posterior.

2. Each anterior element is matched with the closest posterior element measured by sort_by.

Value

An object of the same type as data.

Examples

# See vignette("conjecture") for more examples

conjecture(comms, timestamp, type, "send")

---

Automatically cluster 1-dimensional continuous data.

Description

Automatically cluster 1-dimensional continuous data.

Usage

kluster(x, bw = "SJ", fixed = FALSE)

Arguments

x	Vector to be clustered. Must contain at least 1 non-missing value.
bw	kernel bandwidth. Default "SJ" should suffice more application, however you can supply a custom numeric value. See ?stats::density for more information.
fixed	logical; if TRUE, performs simple 1-dimensional clustering without kernel density estimation. default FALSE.

Value

An integer vector identifying the cluster corresponding to each element in x.

Examples

# Below vector clearly has 2 groups.
# kluster will identify these groups using kernel density estimation.
kluster(c(0.1, 0.2, 1))

# kluster shines in cases where manually assigning groups via "eyeballing" is impractical.
# Suppose we obtained vector 'x' without knowing how it was generated.
set.seed(1)
nodes <- runif(10, min = 0, max = 100)
x <- lapply(nodes, function(x) rnorm(10, mean = x, sd = 0.1))
x <- unlist(x)

kluster(x) # kluster reveals the natural grouping

kluster(x, bw = 10) # adjust bandwidth depending on application

# Example with faithful dataset
faithful$k <- kluster(faithful$eruptions)
library(ggplot2)
ggplot(faithful, aes(eruptions)) +
  geom_density() +
  geom_rug(aes(color = factor(k))) +
  theme_minimal() +
  scale_color_discrete(name = "k")

---

2020 New York Times Headlines

Description

Includes selected headlines and additional metadata for NYT articles throughout 2020. This dataset is not a comprehensive account of all major events from 2020.

Usage

nyt2020

Format

A data frame with 1,830 rows and 6 variables:

headline

Article Headline

abstract

Brief summary of article

byline

Contributing Writers

pub_date

Date of Publication

section_name

NYT section in which article was published

web_url

Article URL

...

Source

Obtained using NYT Developer Portal (Archive API)

---

Augmented data frame filtering.

Description

Imagine dplyr::filter that includes neighboring observations. Choose how many observations to include by adjusting inputs sift.col and scope.

Usage

sift(.data, sift.col, scope, ...)

Arguments

.data	A data frame.
sift.col	Column name, as symbol, to serve as "sifting/augmenting" dimension. Must be non-missing and coercible to numeric.
scope	Specifies augmentation bandwidth relative to "key" observations. Parameter should share the same scale as sift.col. If length 1, bandwidth used is +/- scope. If length 2, bandwidth used is (-scope[1], +scope[2]).
...	Expressions passed to dplyr::filter, of which the results serve as the "key" observations. The same data-masking rules used in dplyr::filter apply here.

Details

sift() can be understood as a 2-step process:

1. .data is passed to dplyr::filter, using subsetting expression(s) provided in .... We'll refer to these intermediate results as "key" observations.

2. For each key observation, sift expands the row selection bidirectionally along dimension specified by sift.col. Any row from the original dataset within scope units of a key observation is captured in the final result.

Essentially, this allows us to "peek" at neighboring rows surrounding the key observations.

Value

A sifted data frame, with 2 additional columns:

• .cluster <int>: Identifies resulting group formed by each key observation and its neighboring rows. When the key observations are close enough together, the clusters will overlap.

• .key <lgl>: TRUE indicates key observation.

Examples

# See current events from same timeframe as 2020 Utah Monolith discovery.
sift(nyt2020, pub_date, scope = 2, grepl("Monolith", headline))

# or Biden's presidential victory.
sift(nyt2020, pub_date, scope = 2, grepl("Biden is elected", headline))

# We can specify lower & upper scope to see what happened AFTER Trump tested positive.
sift(nyt2020, pub_date, scope = c(0, 2), grepl("Trump Tests Positive", headline))

# sift recognizes dplyr group specification.
library(dplyr)
library(mopac)
express %>%
 group_by(direction) %>%
 sift(time, 30, plate == "EAS-1671") # row augmentation performed within groups.

---

Fragments of US & UK population & leaders

Description

These datasets are intended to demonstrate usage of sift::break_join.

Usage

us_uk_pop

us_uk_leaders

Source

See tidyr::who and ggplot2::presidential.

---