Package 'mvalpha'

Published Examples

Description

These data represent examples found in the original paper describing the calculation of multi-valued Krippendorff's alpha by Krippendorff and Craggs (2016).

Usage

ex_table3

ex_table8a

ex_table8b

ex_table8c

ex_table9a

ex_table9b

ex_table9c

Format

Each is a multi-valued nominal set with observers as columns and units as rows.

Source

doi:10.1080/19312458.2016.1228863

References

Krippendorff K, Craggs R (2016). “The Reliability of Multi-Valued Coding of Data.” Communication Methods and Measures, 10(4), 181–198. doi:10.1080/19312458.2016.1228863.

---

Generate Multi-Valued Data Sets

Description

Generate Multi-Valued Data Sets

Usage

generate_mv_data(
  type = "nominal",
  n_units = 10,
  n_observers = 3,
  n_labels = 5,
  tpr = 0.8,
  fpr = 0.01,
  card_pmf = c(0.3, 0.4, 0.2, 0.1),
  p_missing = 0
)

Arguments

type	Data type. One of "nominal", "ordinal", "interval", or "ratio".
n_units	Number of units (rows) in data.
n_observers	Number of observers (cols) in data.
n_labels	Number of possible labels which could be applied to the data.
tpr	True Positive Rate. To generate the data, first a latent set of labels. tpr describes the probability that a latent label will be identified by each observer.
fpr	False Positive Rate. To generate the data, first a latent set of labels. fpr describes the probability that each observer includes each label not in the latent set.
card_pmf	Probability mass function describing the cardinality of observed label sets. Length should be less than or equal to n_labels.
p_missing	Proportion of observations that are randomly missing, indicated by NA. These are distinct from observations which are made, but the observer applied 0 labels, indicated by NULL.

Value

A matrix with n_units rows and n_observers columns with list elements.

Examples

generate_mv_data()

---

Estimate Multi-Valued Krippendorff's Alpha

Description

mvalpha() calculates Krippendorff's alpha statistic when multi-valued observers are allowed to apply multiple values to an observation.

Usage

mvalpha(data, type = "nominal", verbose = TRUE, n_boot = NULL, n_threads = 1)

Arguments

data	a data frame containing a list column for each observer. Each row represents an observation unit, and each cell contains a vector of 0 to w unique values, where w is the number of unique values found in the data set. NA values are used to represent missing observations and NULL values represent the empty set, {}, of responses.
type	a string describing the data type of the label set. This can be "nominal", "ordinal", "interval", or "ratio" and is used to select the appropriate distance metric.
verbose	a logical value which toggles whether status updates are printed to the console while alpha is being calculated.
n_boot	an integer representing the number of bootstrap estimates to calculate for mvDo. The default, NULL, will not generate additional estimates.
n_threads	an integer describing the number of cores to allocate to parallelization.

Value

An object of class mvalpha

References

Krippendorff K, Craggs R (2016). “The Reliability of Multi-Valued Coding of Data.” Communication Methods and Measures, 10(4), 181–198. doi:10.1080/19312458.2016.1228863.

Examples

library(mvalpha)

### replicate example from Table 3 in Krippendoff and Craggs (2016) with bootstrapped estimates

# View data
ex_table3

# # Estimate alpha
# x <- mvalpha(ex_table3, verbose = TRUE, n_boot = 500)
#
# # View result
# x
#
# # View the unique values observed in the data
# x$values
#
# # View the unique labels used to code the data
# x$labels
#
# # Histogram of bootstrapped estimates
# hist(x$bootstrap_mvalpha)

---

Create new mvalpha class object

Description

Wrapper for creating mvalpha class object.

Usage

new_mvalpha(
  mvalpha,
  type,
  mvDo,
  mvDe,
  bootstrap_mvalpha,
  unique_cardinalities,
  units,
  observers,
  labels,
  values,
  values_by_unit,
  dist_CK,
  p_CK,
  data
)

Arguments

mvalpha	Multi-valued alpha estimate
type	a string describing the data type of the label set. This can be "nominal", "ordinal", "interval", or "ratio" and is used to select the appropriate distance metric.
mvDo	Observed disagreement
mvDe	Expected disagreement
bootstrap_mvalpha	Bootstrap estimates of mvalpha
unique_cardinalities	Numeric vector of the unique cardinalities observed in the data
units	Names of units
observers	Names of observers
labels	Unique labels used in data
values	Unique values used in data
values_by_unit	Table of values by unit
dist_CK	Distance matrix for label sets C and K
p_CK	Probability matrix for label sets C and K
data	a data frame containing a list column for each observer. Each row represents an observation unit, and each cell contains a vector of 0 to w unique values, where w is the number of unique values found in the data set. NA values are used to represent missing observations and NULL values represent the empty set, {}, of responses.

Value

an mvalpha object

---

Print mvalpha class object

Description

Print generic

Usage

## S3 method for class 'mvalpha'
print(x, ...)

Arguments

x	mvalpha object
...	additional parameters

Value

invisibly returns the alpha estimate of an mvalpha object

---

Efficient Set Operations

Description

Find the intersection and set difference(s) of two sets all at once and more efficiently than calling base::intersect() and base::setdiff() separately. Based on this stackoverflow answer https://stackoverflow.com/a/72631719

Usage

set_ops(A, B, type)

Arguments

A, B	sets (vectors) of elements
type	a string describing the data type of the label set. This can be "nominal", "ordinal", "interval", or "ratio" and is used to select the appropriate distance metric.

---