Title:	Data Modification and Analysis for Communication Research
Version:	0.4.1
Description:	Provides convenience functions for common data modification and analysis tasks in communication research. This includes functions for univariate and bivariate data analysis, index generation and reliability computation, and intercoder reliability tests. All functions follow the style and syntax of the tidyverse, and are construed to perform their computations on multiple variables at once. Functions for univariate and bivariate data analysis comprise summary statistics for continuous and categorical variables, as well as several tests of bivariate association including effect sizes. Functions for data modification comprise index generation and automated reliability analysis of index variables. Functions for intercoder reliability comprise tests of several intercoder reliability estimates, including simple and mean pairwise percent agreement, Krippendorff's Alpha (Krippendorff 2004, ISBN: 9780761915454), and various Kappa coefficients (Brennan & Prediger 1981 <doi:10.1177/001316448104100307>; Cohen 1960 <doi:10.1177/001316446002000104>; Fleiss 1971 <doi:10.1037/h0031619>).
License:	GPL-3
URL:	https://joon-e.github.io/tidycomm/
BugReports:	https://github.com/joon-e/tidycomm/issues
Depends:	R (≥ 2.10)
Imports:	car, dplyr, fastDummies, forcats, GGally, ggplot2, glue, lm.beta, lubridate, magrittr, MASS, MBESS, misty, pillar, purrr, rlang, stringr, tibble, tidyr, tidyselect
Suggests:	covr, knitr, rmarkdown, testthat (≥ 2.1.0)
VignetteBuilder:	knitr
Encoding:	UTF-8
LazyData:	true
RoxygenNote:	7.2.3
NeedsCompilation:	no
Packaged:	2024-02-22 11:57:03 UTC; LaraK
Author:	Julian Unkel [aut, cre], Mario Haim [aut], Lara Kobilke [aut]
Maintainer:	Julian Unkel <julian.unkel@gmail.com>
Repository:	CRAN
Date/Publication:	2024-02-22 12:20:02 UTC

Pipe operator

Description

See magrittr::%>% for details.

Usage

lhs %>% rhs

Worlds of Journalism sample data

Description

A subset of data from the Worlds of Journalism 2012-16 study containing survey data of 1,200 journalists from five European countries.

Usage

WoJ

Format

A data frame with 1200 rows and 15 variables:

country

Country of residence

reach

Reach of medium

employment

Current employment situation

temp_contract

Type of contract (if current employment situation is either full-time or part-time

autonomy_selection

Autonomy in news story selection, scale from 1 (no freedom at all) to 5 (complete freedom)

autonomy_emphasis

Autonomy in news story emphasis, scale from 1 (no freedom at all) to 5 (complete freedom)

ethics_1

Agreement with statement "Journalists should always adhere to codes of professional ethics, regardless of situation and context", scale from 1 (strongly disagree) to 5 (strongly agree) (reverse-coded!)

ethics_2

Agreement with statement "What is ethical in journalism depends on the specific situation.", scale from 1 (strongly disagree) to 5 (strongly agree)

ethics_3

Agreement with statement "What is ethical in journalism is a matter of personal judgment.", scale from 1 (strongly disagree) to 5 (strongly agree)

ethics_4

Agreement with statement "It is acceptable to set aside moral standards if extraordinary circumstances require it.", scale from 1 (strongly disagree) to 5 (strongly agree)

work_experience

Work experience as a journalist in years

trust_parliament

Trust placed in parliament, scale from 1 (no trust at all) to 5 (complete trust)

trust_government

Trust placed in government, scale from 1 (no trust at all) to 5 (complete trust)

trust_parties

Trust placed in parties, scale from 1 (no trust at all) to 5 (complete trust)

trust_politicians

Trust placed in politicians in general, scale from 1 (no trust at all) to 5 (complete trust)

Source

‘⁠https://worldsofjournalism.org/data/data-and-key-tables-2012-2016⁠’

Add index

Description

Add a rowwise mean or sum index of specific variables to the dataset.

Usage

add_index(data, name, ..., type = "mean", na.rm = TRUE, cast.numeric = FALSE)

Arguments

Value

a tdcmm model

See Also

get_reliability() to compute reliability estimates of added index variables.

Examples

WoJ %>% add_index(ethical_flexibility, ethics_1, ethics_2, ethics_3, ethics_4)
WoJ %>% add_index(ethical_flexibility, ethics_1, ethics_2, ethics_3, ethics_4, type = "sum")

Categorize numeric variables into categories

Description

This function recodes one or more numeric variables into categorical variables based on a specified lower end, upper end, and intermediate breaks. The intervals created include the right endpoint of the interval. For example, breaks = c(2, 3) with lower_end = 1 and upper_end = 5 creates intervals from 1 to <= 2, >2 to <= 3, and >3 to <= 5. If the lower or upper ends are not provided, the function defaults to the minimum and maximum values of the data and issues a warning. This default behavior is prone to errors, however, because a scale may not include its actual lower and upper ends which might in turn affect the recoding process. Hence, it is strongly suggested to manually set the lower and upper bounds of the original continuous scale.

Usage

categorize_scale(
  data,
  ...,
  breaks,
  labels,
  lower_end = NULL,
  upper_end = NULL,
  name = NULL,
  overwrite = FALSE
)

Arguments

Value

A modified tibble or tdcmm model with the recoded variables.

See Also

Other scaling: center_scale(), dummify_scale(), minmax_scale(), recode_cat_scale(), reverse_scale(), setna_scale(), z_scale()

Examples

WoJ %>%
dplyr::select(trust_parliament, trust_politicians) %>%
categorize_scale(trust_parliament, trust_politicians,
lower_end = 1, upper_end = 5, breaks = c(2, 3),
labels = c("Low", "Medium", "High"), overwrite = FALSE)
WoJ %>%
dplyr::select(autonomy_selection) %>%
categorize_scale(autonomy_selection, breaks = c(2, 3, 4),
lower_end = 1, upper_end = 5,
labels = c("Low", "Medium", "High", "Very High"),
name = "autonomy_in_categories")

Center numeric, continuous variables

Description

This function centers the specified numeric columns or all numeric columns if none are specified. A centered scale has a mean of 0.0.

Usage

center_scale(data, ..., name = NULL, overwrite = FALSE)

Arguments

Value

A tdcmm model with the centered variable(s).

See Also

Other scaling: categorize_scale(), dummify_scale(), minmax_scale(), recode_cat_scale(), reverse_scale(), setna_scale(), z_scale()

Examples

WoJ %>% dplyr::select(autonomy_emphasis) %>% center_scale(autonomy_emphasis)
WoJ %>% center_scale(autonomy_emphasis, name = "my_centered_variable")
WoJ %>% center_scale(overwrite = TRUE)
WoJ %>%
  center_scale(autonomy_emphasis) %>%
  tab_frequencies(autonomy_emphasis, autonomy_emphasis_centered)

Compute correlation coefficients

Description

Computes correlation coefficients for all combinations of the specified variables. If no variables are specified, all numeric (integer or double) variables are used.

Usage

correlate(data, ..., method = "pearson", partial = NULL, with = NULL)

Arguments

Value

a tdcmm model

Examples

WoJ %>% correlate(ethics_1, ethics_2, ethics_3)
WoJ %>% correlate()
WoJ %>% correlate(ethics_1, ethics_2, ethics_3, with = work_experience)
WoJ %>% correlate(autonomy_selection, autonomy_emphasis, partial = work_experience)
WoJ %>% correlate(with = work_experience)

Crosstab variables

Description

Computes contingency table for one independent (column) variable and one or more dependent (row) variables.

Usage

crosstab(
  data,
  col_var,
  ...,
  add_total = FALSE,
  percentages = FALSE,
  chi_square = FALSE
)

Arguments

Value

a tdcmm model

See Also

Other categorical: tab_frequencies()

Examples

WoJ %>% crosstab(reach, employment)
WoJ %>% crosstab(reach, employment, add_total = TRUE, percentages = TRUE, chi_square = TRUE)

Describe numeric variables

Description

Describe numeric variables by several measures of central tendency and variability. If no variables are specified, all numeric (integer or double) variables are described.

Usage

describe(data, ..., na.rm = TRUE)

Arguments

Details

• N: number of valid cases (i.e., all but missing)

• Missing: number of NA cases

• M: mean average

• SD: standard deviation, sd

• Min: minimum value, min

• Q25: 25% quantile, quantile

• Mdn: median average, same as 50% quantile

• Q75: 75% quantile, quantile

• Max: maximum value, max

• Range: difference between Min and Max

• CI_95_LL: M - Q(0.975) \times \frac{SD}{\sqrt{N}} where Q(0.975) denotes Student t's stats::quantile function with a probability of 0.975 and N-1 degrees of freedom

• CI_95_UL: M + Q(0.975) \times \frac{SD}{\sqrt{N}} where Q(0.975) denotes Student t's stats::quantile function with a probability of 0.975 and N-1 degrees of freedom

• Skewness: traditional Fisher-Pearson coefficient of skewness of valid cases as per \frac{\frac{1}{N} \sum\limits_{i=1}^N (x_{i}-\overline{x})^3}{[\frac{1}{N}\sum\limits_{i=1}^N (x_{i}-\overline{x})^2]^{3/2}} where \overline{x} denotes M, following Doane & Seward (2011, p. 6, 1a). See DOI doi:10.1080/10691898.2011.11889611.

• Kurtosis: empirical sample kurtosis (i.e., standardized fourth population moment about the mean) as per \frac{\sum (x-\overline{x})^4 / N}{(\sum (x-\overline{x})^2 / N)^2}, following DeCarlo (1997, p. 292, b2). See DOI doi:10.1037/1082-989X.2.3.292.

Value

a tdcmm model

See Also

Other descriptives: describe_cat(), tab_percentiles()

Examples

WoJ %>% describe(autonomy_selection, autonomy_emphasis, work_experience)
fbposts %>% describe(n_pictures)

Describe categorical variables

Description

Describe categorical variables by N, number of unique values, and mode. Note that in case of multiple modes, the first mode by order of values is chosen.

Usage

describe_cat(data, ...)

Arguments

Details

If no variables are specified, all categorical (character or factor) variables are described.

• N: number of valid cases (i.e., all but missing)

• Missing: number of NA cases

• Unique: number of unique categories in a given variable, without Missing

• Mode: mode average (if multiple modes exist, first mode by order of values is returned)

• Mode_N: number of cases reflecting the Mode

Value

a tdcmm model

See Also

Other descriptives: describe(), tab_percentiles()

Examples

WoJ %>% describe_cat(reach, employment, temp_contract)
fbposts %>% describe_cat(type)

Gray design

Description

Gray design

Usage

design_gray()

Value

a list with main_color_1, a vector of 12 main_colors, a corresponding main_contrast_1 (the color of text to write on top of the main color) and a corresponding main_contrasts, the main_size (for lines), a comparison_linetype, comparison_color, and comparison_size for all lines that act as comparative lines, and a ggplot2 theme

Grey design

Description

Grey design

Usage

design_grey()

Value

a list with main_color_1, a vector of 12 main_colors, a corresponding main_contrast_1 (the color of text to write on top of the main color) and a corresponding main_contrasts, the main_size (for lines), a comparison_linetype, comparison_color, and comparison_size for all lines that act as comparative lines, and a ggplot2 theme

Colorbrewer-inspired design with focus on LMU (lmu.de) green

Description

Colorbrewer-inspired design with focus on LMU (lmu.de) green

Usage

design_lmu()

Value

a list with main_color_1, a vector of 12 main_colors, a corresponding main_contrast_1 (the color of text to write on top of the main color) and a corresponding main_contrasts, the main_size (for lines), a comparison_linetype, comparison_color, and comparison_size for all lines that act as comparative lines, and a ggplot2 theme

Convert categorical variables to dummy variables

Description

This function transforms specified categorical variables into dummy variables. Each level of the categorical variable is represented by a new dummy variable. Missing values are retained. These new dummy variables are appended to the original data frame. This function does not allow specifying new column names for the dummy variables. Instead, it follows a consistent naming pattern: the new dummy variables are named using the original variable name with the category value appended. For example, if a categorical variable named "autonomy" with levels "low", "medium", "high" is dummified, the new dummy variables will be named "autonomy_low", "autonomy_medium", "autonomy_high".

Usage

dummify_scale(data, ..., overwrite = FALSE)

Arguments

Value

A tdcmm model with the dummy variables appended.

See Also

Other scaling: categorize_scale(), center_scale(), minmax_scale(), recode_cat_scale(), reverse_scale(), setna_scale(), z_scale()

Examples

WoJ %>% dplyr::select(temp_contract) %>% dummify_scale(temp_contract)
WoJ %>% categorize_scale(autonomy_emphasis, breaks = c(2, 3),
labels = c('low', 'medium', 'high')) %>%
dummify_scale(autonomy_emphasis_cat) %>% dplyr::select(starts_with('autonomy_emphasis'))

Expand a range with a multiplicative or additive constant

Description

Expand a range with a multiplicative or additive constant

Usage

expand_range(range, mul = 0, add = 0, zero_width = 1)

Arguments

Facebook posts reliability test

Description

45 political facebook posts coded by 6 coders for an intercoder reliability test, focused on populist messages.

Usage

fbposts

Format

A data frame with 270 rows and 7 variables

post_id

Numeric id of the coded Facebook post

coder_id

Numeric id of the coder

type

Type of Facebook post, one of "link", "photo", "status", or "video

n_pictures

Amount of pictures attached to the post, ranges from 0 to 6

pop_elite

Populism indicator: Does the Facebook post attack elites?, 0 = "no attacks on elites", 1 = "attacks political actors", 2 = "attacks public administration actors", 3 = "attacks economical actors", 4 = "attacks media actors/journalists", 9 = "attacks other elites"

pop_people

Populism indicator: Does the Facebook refer to 'the people'?, 0 = "does not refer to 'the people'", 1 = "refers to 'the people'"

pop_othering

Populism indicator: Does the Facebook attack 'others'?, 0 = "no attacks on 'others'", 1 = "attacks other cultures", 2 = "attacks other political stances", 3 = "attacks other 'others'"

Get reliability estimates of index variables

Description

Get reliability estimates of index variables created with add_index.

Usage

get_reliability(
  data,
  ...,
  type = "alpha",
  interval.type = NULL,
  bootstrap.samples = NULL,
  conf.level = NULL,
  progress = FALSE
)

Arguments

Value

a tdcmm model

See Also

add_index() to create index variables

Examples

WoJ %>%
  add_index(ethical_flexibility, ethics_1, ethics_2, ethics_3, ethics_4) %>%
  get_reliability()

Incivil Comments Data

Description

A dataset of a preregistered factorial survey experiment with a nationally representative sample of 964 German online users. Participants were presented with manipulated user comments that included statements associated with incivil discourse (such as profanity and attacks on arguments) and intolerant discourse (such as offensive stereotyping and violent threats). Participants rated the comments, e.g. offensiveness, harm to society, and their intention to delete the comment containing the statement.

Usage

incvlcomments

Format

A data frame of 3856 observations nested in 964 participants and 22 variables:

participant_num

Numeric id of the participant

age

Age of the participant

male

Gender of the participant, either 'male' or 'not male'

high_education

Level of formal education of the participant, either 'high formal education' or 'low formal education'

comment_num

Numeric id of the comment that the participant was exposed to

issue

The subject of the comment that the participant was exposed to, either 'Gender', 'Abortion', 'Climate', or 'Migration'

profanity

Whether the comment contained profanities as an indicator of incivility

attacks_argument

Whether the comment contained attacks towards arguments as an indicator of incivility

offensive_stereotyping

Whether the comment contained offensive stereotypes as an indicator of intolerant discourse

violent_threats

Whether the comment contained violent threats as an indicator of intolerant discourse

offensiveness

Rate statement whether the comment is being perceived as offensive & hostile (Scale from 1 to 7)

adequacy

Rate statement whether the comment is being perceived as necessary & accurate (Scale from 1 to 7)

harm_to_society

Rate statement whether the comment is being perceived as harmful to society (Scale from 1 to 7)

deletion_intention

Whether the participant wants to delete the comment

similarity_poster

How similar the participant feels to the person who created the post (Scale from 1 to 7)

similarity_group

How similar the participant feels to the group of people criticized in the post (Scale from 1 to 7)

attitude_gender

Rate agreement with statements on gender policies (Scale from 1 to 7)

attitude_abortion

Rate agreement with statements on abortion (Scale from 1 to 7)

attitude_migration

Rate agreement with statements on migration (Scale from 1 to 7)

attitude_climate

Rate agreement with statements on climate change (Scale from 1 to 7)

left_right_placement

Placement on a political spectrum from left to right (Scale from 1 to 9)

freedom_of_speech

Rate agreement with statements about the freedom of speech and expression (Scale from 1 to 7)

Details

The dataset was created from the OSF project: Differential perceptions of and reactions to incivil and intolerant user comments, corresponding to the paper: Kümpel, A. S., Unkel, J (2023). Differential perceptions of and reactions to incivil and intolerant user comments, Journal of Computer-Mediated Communication, Volume 28, Issue 4, https://doi.org/10.1093/jcmc/zmad018

Source

https://osf.io/w92vj

Rescale numeric continuous variables to new minimum/maximum boundaries

Description

Given a specified minimum and maximum, this function translates each value into a new value within this specified range. The transformation maintains the relative distances between values, resulting in changes to the mean and standard deviations. However, if both the original scale and the transformed scale are z-standardized, they will be equal again, indicating that the relative positions and distributions of the values remain consistent.

Usage

minmax_scale(
  data,
  ...,
  change_to_min = 0,
  change_to_max = 1,
  name = NULL,
  overwrite = FALSE
)

Arguments

Value

A tdcmm model with the min-max scaled variable(s).

See Also

Other scaling: categorize_scale(), center_scale(), dummify_scale(), recode_cat_scale(), reverse_scale(), setna_scale(), z_scale()

Examples

WoJ %>% minmax_scale(autonomy_emphasis, change_to_min = 0,
change_to_max = 1)
WoJ %>% minmax_scale(autonomy_emphasis, name = "my_scaled_variable",
change_to_min = 0, change_to_max = 1)
WoJ %>%
  minmax_scale(autonomy_emphasis, change_to_min = 0, change_to_max = 1) %>%
  tab_frequencies(autonomy_emphasis, autonomy_emphasis_0to1)

Access model(s) used to estimate output

Description

Returns model objects used to estimate tdcmm output.

Usage

model(x, ...)

Arguments

Value

A model object or a list of model objects

tdcmm output constructor

Description

Creates a new tdcmm class output object.

The tdcmm class is a subclass of a tbl_df, also know as a "tibble", used for augmenting output tibbles of tidycomm functions with additional information. For output based on statistical tests, the model object(s) estimated and any performed assumption checks.

tdcmm objects in tidycomm are further subclassed with individual classes handling visualization and printing per "output" type.

Usage

new_tdcmm(x, func, data, model = NULL, checks = NULL, params = list())

is_tdcmm(x)

Arguments

Functions

• is_tdcmm(): Test for class tdcmm

Out of bounds handling

Description

This set of functions modify data values outside a given range. The ⁠oob_*()⁠ functions are designed to be passed as the oob argument of ggplot2 continuous and binned scales, with oob_discard being an exception.

These functions affect out of bounds values in the following ways:

• oob_censor() replaces out of bounds values with NAs. This is the default oob argument for continuous scales.

• oob_censor_any() acts like oob_censor(), but also replaces infinite values with NAs.

• oob_squish() replaces out of bounds values with the nearest limit. This is the default oob argument for binned scales.

• oob_squish_any() acts like oob_squish(), but also replaces infinite values with the nearest limit.

• oob_squish_infinite() only replaces infinite values by the nearest limit.

• oob_keep() does not adjust out of bounds values. In position scales, behaves as zooming limits without data removal.

• oob_discard() removes out of bounds values from the input. Not suitable for ggplot2 scales.

Usage

oob_censor(x, range = c(0, 1), only.finite = TRUE)

oob_censor_any(x, range = c(0, 1))

oob_discard(x, range = c(0, 1))

oob_squish(x, range = c(0, 1), only.finite = TRUE)

oob_squish_any(x, range = c(0, 1))

oob_squish_infinite(x, range = c(0, 1))

oob_keep(x, range = c(0, 1))

censor(x, range = c(0, 1), only.finite = TRUE)

discard(x, range = c(0, 1))

squish(x, range = c(0, 1), only.finite = TRUE)

squish_infinite(x, range = c(0, 1))

Arguments

Details

The oob_censor_any() and oob_squish_any() functions are the same as oob_censor() and oob_squish() with the only.finite argument set to FALSE.

Replacing position values with NAs, as oob_censor() does, will typically lead to removal of those datapoints in ggplot.

Setting ggplot coordinate limits is equivalent to using oob_keep() in position scales.

Value

Most oob_() functions return a vector of numerical values of the same length as the x argument, wherein out of bounds values have been modified. Only oob_discard() returns a vector of less than or of equal length to the x argument.

Old interface

censor(), squish(), squish_infinite() and discard() are no longer recommended; please use oob_censor(), oob_squish(), oob_squish_infinite() and oob_discard() instead.

Author(s)

oob_squish(): Homer Strong homer.strong@gmail.com

Helper function for labelling purposes

Description

Helper function for labelling purposes

Usage

percentage_labeller(x)

Arguments

Value

a string with formatted % (rounded and suffixed)

Recode one or more categorical variables into new categories

Description

This function transforms one or more categorical variables into new categories based on specified mapping. For unmatched cases not specified in the mapping, a default value can be assigned. Missing values are retained.

Usage

recode_cat_scale(
  data,
  ...,
  assign = NULL,
  other = NA,
  overwrite = FALSE,
  name = NULL
)

Arguments

Value

A tdcmm model or a tibble.

See Also

Other scaling: categorize_scale(), center_scale(), dummify_scale(), minmax_scale(), reverse_scale(), setna_scale(), z_scale()

Examples

WoJ %>%
recode_cat_scale(country,
assign = c("Germany" = 1, "Switzerland" = 2), overwrite = TRUE)
WoJ %>%
recode_cat_scale(country,
assign = c("Germany" = "german", "Switzerland" = "swiss"), other = "other",
overwrite = TRUE)
WoJ %>%
recode_cat_scale(ethics_1, ethics_2,
assign = c(`1` = 5, `2` = 4, `3` = 3, `4` = 2, `5` = 1), other = 6, overwrite = TRUE)
WoJ %>%
recode_cat_scale(ethics_1, ethics_2,
assign = c(`1` = "very low", `2` = "low", `3` = "medium", `4` = "high", `5` = "very high"),
overwrite = TRUE)
WoJ %>%
dplyr::select(temp_contract) %>% recode_cat_scale(temp_contract,
assign = c(`Permanent` = "P", `Temporary` = "T"), other = "O")

Compute linear regression

Description

Computes linear regression for all independent variables on the specified dependent variable. Linear modeling of multiple independent variables uses stepwise regression modeling. If specified, preconditions for (multi-)collinearity and for homoscedasticity are checked.

Usage

regress(
  data,
  dependent_var,
  ...,
  check_independenterrors = FALSE,
  check_multicollinearity = FALSE,
  check_homoscedasticity = FALSE
)

Arguments

Value

a tdcmm model

Examples

WoJ %>% regress(autonomy_selection, ethics_1)
WoJ %>% regress(autonomy_selection, work_experience, trust_government)

Rescale continuous vector to have specified minimum and maximum

Description

Rescale continuous vector to have specified minimum and maximum

Usage

rescale(x, to, from, ...)

## S3 method for class 'numeric'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE, finite = TRUE), ...)

## S3 method for class 'dist'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE, finite = TRUE), ...)

## S3 method for class 'logical'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE, finite = TRUE), ...)

## S3 method for class 'POSIXt'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE, finite = TRUE), ...)

## S3 method for class 'Date'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE, finite = TRUE), ...)

## S3 method for class 'integer64'
rescale(x, to = c(0, 1), from = range(x, na.rm = TRUE), ...)

Arguments

Rescale numeric vector to have specified maximum

Description

Rescale numeric vector to have specified maximum

Usage

rescale_max(x, to = c(0, 1), from = range(x, na.rm = TRUE))

Arguments

Rescale vector to have specified minimum, midpoint, and maximum

Description

Rescale vector to have specified minimum, midpoint, and maximum

Usage

rescale_mid(x, to, from, mid, ...)

## S3 method for class 'numeric'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid = 0, ...)

## S3 method for class 'logical'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid = 0, ...)

## S3 method for class 'dist'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid = 0, ...)

## S3 method for class 'POSIXt'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid, ...)

## S3 method for class 'Date'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid, ...)

## S3 method for class 'integer64'
rescale_mid(x, to = c(0, 1), from = range(x, na.rm = TRUE), mid = 0, ...)

Arguments

Don't perform rescaling

Description

Don't perform rescaling

Usage

rescale_none(x, ...)

Arguments

Reverse numeric, logical, or date/time continuous variables

Description

Reverses a continuous scale into a new variable. A 5-1 scale thus turns into a 1-5 scale. Missing values are retained. For a given continuous variable the lower and upper end of the scale should be provided. If they are not provided, the function assumes the scale's minimum and maximum value to represent these lower/upper ends (and issues a warning about this fact). This default behavior is prone to errors, however, because a scale may not include its actual lower and upper ends which might in turn affect correct reversing. Hence, it is strongly suggested to manually set the lower and upper bounds of the original continuous scale.

Usage

reverse_scale(
  data,
  ...,
  lower_end = NULL,
  upper_end = NULL,
  name = NULL,
  overwrite = FALSE
)

Arguments

Value

A tdcmm model with the reversed variable(s).

See Also

Other scaling: categorize_scale(), center_scale(), dummify_scale(), minmax_scale(), recode_cat_scale(), setna_scale(), z_scale()

Examples

WoJ %>% reverse_scale(autonomy_emphasis, lower_end = 0, upper_end = 1)
WoJ %>% reverse_scale(autonomy_emphasis, name = "my_reversed_variable",
lower_end = 0, upper_end = 1)
WoJ %>% reverse_scale(overwrite = TRUE)
WoJ %>%
  reverse_scale(autonomy_emphasis, lower_end = 0, upper_end = 1) %>%
  tab_frequencies(autonomy_emphasis, autonomy_emphasis_rev)

Set specified values to NA in selected variables or entire data frame

Description

This function allows users to set specific values to NA in chosen variables within a data frame. It can handle numeric, character, and factor variables.

Usage

setna_scale(data, ..., value, name = NULL, overwrite = FALSE)

Arguments

Value

A tdcmm model or a tibble.

See Also

Other scaling: categorize_scale(), center_scale(), dummify_scale(), minmax_scale(), recode_cat_scale(), reverse_scale(), z_scale()

Examples

WoJ %>%
dplyr::select(autonomy_emphasis) %>%
setna_scale(autonomy_emphasis, value = 5)
WoJ %>%
dplyr::select(autonomy_emphasis) %>%
setna_scale(autonomy_emphasis, value = 5, name = "new_na_autonomy")
WoJ %>%
setna_scale(value = c(2, 3, 4), overwrite = TRUE)
WoJ %>%
dplyr::select(country) %>% setna_scale(country, value = "Germany")
WoJ %>%
dplyr::select(country) %>% setna_scale(country, value = c("Germany", "Switzerland"))

SNS Comments data

Description

A dataset of 630 German participants in an online experiment. The experiment investigated the effects of user comments on social network sites (SNS) on individuals' perceptions of journalistic quality. The researchers varied the subject of the article (factor 1: 'Copyright directive' or 'Social housing'), the order of comment presentation (factor 2: before or after the article) and the valence of the comments (factor 3: positive or negative).

Usage

snscomments

Format

A data frame of 630 observations and 15 variables:

age

Age of the participant

gender

Gender of the participant, either 'female' or 'not female'

education

Level of formal education of the participant, either 'low formal education' or 'high formal education'

need_cognition

Index measuring the psychological trait of a person to enjoy thinking, calculated from several survey items

prior_knowledge

Index measuring a person's prior knowledge of the presented subject of the article, calculated from several survey items

group

Numeric id of the group that the participant was in during the experiment

issue

Subject of the article that the participant was given to read, either 'Copyright directive' or 'Social housing'

order

Order of the comments that the participant was exposed to, either 'Comments after', 'Comments before', or 'Control group'

valence

Valence of the comments that the participant was exposed to, either 'Negative', 'Positive', or 'Control group'

control_group

Indicates whether the participant was in the 'Control group' or 'Experimental group'

medium_evaluation

Index measuring participant's evaluation of the medium's quality, calculated from several survey items

article_evaluation

Index measuring participant's evaluation of the article's quality, calculated from several survey items

comments_quality

Participant's perception of the quality of the comments

comments_valence

Participant's perception of the valence of the comments

article_elaboration

Participant's measure of how much attention they put in reading the article

Details

This dataset was created from the OSF project: https://osf.io/r867v/, corresponding to the paper: Kümpel, A. S., & Unkel, J. (2020). Negativity wins at last: How presentation order and valence of user comments affect perceptions of journalistic quality. Journal of Media Psychology: Theories, Methods, and Applications, 32(2), 89–99. doi:10.1027/1864-1105/a000261

Source

https://osf.io/r867v/

Compute t-tests

Description

Computes t-tests for one group variable and specified test variables. If no variables are specified, all numeric (integer or double) variables are used. A Levene's test will automatically determine whether the pooled variance is used to estimate the variance. Otherwise the Welch (or Satterthwaite) approximation to the degrees of freedom is used.

Usage

t_test(
  data,
  group_var,
  ...,
  var.equal = TRUE,
  paired = FALSE,
  pooled_sd = TRUE,
  levels = NULL,
  case_var = NULL,
  mu = NULL
)

Arguments

Value

a tdcmm model

Examples

WoJ %>% t_test(temp_contract, autonomy_selection, autonomy_emphasis)
WoJ %>% t_test(temp_contract)
WoJ %>% t_test(employment, autonomy_selection, autonomy_emphasis,
  levels = c("Full-time", "Freelancer"))
WoJ %>% t_test(autonomy_selection, mu = 3.62)

Tabulate frequencies

Description

Tabulates frequencies for one or more categorical variable, including relative, and cumulative frequencies.

Usage

tab_frequencies(data, ...)

Arguments

Value

a tdcmm model

See Also

Other categorical: crosstab()

Examples

WoJ %>% tab_frequencies(employment)
WoJ %>% tab_frequencies(employment, country)

Tabulate percentiles for numeric variables

Description

This function tabulates specified percentiles for given numeric variables. If no variables are provided, the function will attempt to describe all numeric (either integer or double) variables found within the input. The percentiles are calculated based on the levels parameter, which defaults to every 10% from 10% to 90%. NA values are always removed because the concept of a percentile is based on ranking. As NA is not a value, it cannot be ordered in relation to actual numbers.

Usage

tab_percentiles(
  data,
  ...,
  levels = c(0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1)
)

Arguments

Value

a tdcmm model

See Also

Other descriptives: describe_cat(), describe()

Examples

WoJ %>% tab_percentiles(work_experience)
WoJ %>% tab_percentiles(work_experience, autonomy_emphasis)

tdcmm class

Description

The tdcmm class is a specialized type of data frame that is a subclass of tbl_df, also known as a "tibble". It is designed to be used with tidycomm functions to provide additional information and context to the output tibbles generated by these functions. This subclass specifically augments output based on statistical tests, by including the relevant model object(s) estimated during the analysis, as well as any performed assumption checks.

tdcmm objects in tidycomm are further subclassed with individual classes handling visualization and printing per "output" type.

Perform an intercoder reliability test

Description

Performs an intercoder reliability test by computing various intercoder reliability estimates for the included variables

Usage

test_icr(
  data,
  unit_var,
  coder_var,
  ...,
  levels = NULL,
  na.omit = FALSE,
  agreement = TRUE,
  holsti = TRUE,
  kripp_alpha = TRUE,
  cohens_kappa = FALSE,
  fleiss_kappa = FALSE,
  brennan_prediger = FALSE,
  lotus = FALSE,
  s_lotus = FALSE
)

Arguments

Value

a tdcmm model

References

Brennan, R. L., & Prediger, D. J. (1981). Coefficient Kappa: Some uses, misuses, and alternatives. Educational and Psychological Measurement, 41(3), 687-699. https://doi.org/10.1177/001316448104100307

Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement, 20(1), 37-46. https://doi.org/10.1177/001316446002000104

Fleiss, J. L. (1971). Measuring nominal scale agreement among many raters. Psychological Bulletin, 76(5), 378-382. https://doi.org/10.1037/h0031619

Fretwurst, B. (2015). Reliabilität und Validität von Inhaltsanalysen. Mit Erläuterungen zur Berechnung des Reliabilitätskoeffizienten „Lotus“ mit SPSS. In W. Wirth, K. Sommer, M. Wettstein, & J. Matthes (Ed.), Qualitätskriterien in der Inhaltsanalyse (S. 176–203). Herbert von Halem.

Krippendorff, K. (2011). Computing Krippendorff's Alpha-Reliability. Retrieved from http://repository.upenn.edu/asc_papers/43

von Eye, A. (2006). An Alternative to Cohen's Kappa. European Psychologist, 11(1), 12-24. https://doi.org/10.1027/1016-9040.11.1.12

Examples

fbposts %>% test_icr(post_id, coder_id, pop_elite, pop_othering)
fbposts %>% test_icr(post_id, coder_id, levels = c(n_pictures = "ordinal"), fleiss_kappa = TRUE)

Create correlation matrix

Description

Turns the tibble exported from correlate into a correlation matrix.

Usage

to_correlation_matrix(data, verbose = FALSE)

Arguments

Value

a tdcmm model

Examples

WoJ %>% correlate() %>% to_correlation_matrix()

Compute one-way ANOVAs

Description

Computes one-way ANOVAs for one group variable and specified test variables. If no variables are specified, all numeric (integer or double) variables are used. A Levene's test will automatically determine whether a classic ANOVA is used. Otherwise Welch's ANOVA with a (Satterthwaite's) approximation to the degrees of freedom is used.

Usage

unianova(data, group_var, ..., descriptives = FALSE, post_hoc = FALSE)

Arguments

Value

a tdcmm model

Examples

WoJ %>% unianova(employment, autonomy_selection, autonomy_emphasis)
WoJ %>% unianova(employment, descriptives = TRUE, post_hoc = TRUE)
## Not run: 
WoJ %>% unianova(employment)

## End(Not run)

Visualize tidycomm output

Description

Returns ggplot2 visualization appropriate to respective tdcmm model (see list below). Returns NULL (and a warning) if no visualization has been implemented for the particular model.

Usage

## S3 method for class 'tdcmm_ctgrcl'
visualize(x, ..., .design = design_lmu())

## S3 method for class 'tdcmm_crrltn'
visualize(x, which = "jitter", ..., .design = design_lmu())

## S3 method for class 'tdcmm_dscrb'
visualize(x, ..., .design = design_lmu())

## S3 method for class 'tdcmm_rgrssn'
visualize(x, which = "jitter", ..., .design = design_lmu())

## S3 method for class 'tdcmm_prcntl'
visualize(x, ..., .design = design_lmu())

visualize(x, ..., .design = design_lmu())

## S3 method for class 'tdcmm_ttst'
visualize(x, ..., .design = design_lmu())

## S3 method for class 'tdcmm_nnv'
visualize(x, ..., .design = design_lmu())

Arguments

Details

• describe(): horizontal box plot depicting a box from Q25 to Q75, a thick line for Mdn, and two whiskers to Min/Max respectively; no additional arguments

• describe_cat(): horizontal bar plot depicting number of occurrences; no additional arguments

• tab_frequencies(): either a histogram (if 1 variable is given) or multiple histograms wrapped, 5+ variables issue a warning about readability; no additional arguments

• tab_percentiles(): quantile plot

• crosstab(): horizontal stacked bar plot, either absolute or relative (depending on the percentages argument in crosstab())

• t_test(): plot with points and appended 95% confidence intervals; no additional arguments

• unianova(): plot with points and appended 95% confidence intervals; no additional arguments

• correlate(): plot as scatter; for more than 2 variables, a correlogram is plotted (just like for to_correlation_matrix()); use the which parameter to select how points are visualized:

  • "jitter" adds a bit of random noise to each point to better reflect categorical values

  • "alpha" depicts points slightly transparent so that multiple points in the same position are more easily visible

• correlate(): for partial correlation, a scatter plot with some jitter is plotted using the residuals between the control variable and (a) the dependent as well as (b) the independent variable; no additional arguments

• to_correlation_matrix(): plot as correlogram building on GGally::ggpairs() with jittered scatter plots in lower half, histograms as diagonals, and correlation coefficients with 95% confidence intervals in upper half

• regress(): plot regression results as scatter (without jitter) and an additional depicted model line with including its 95% confidence intervals; alternatively, visual check inspection helpers can be plotted through the which parameter which can be set to yield one of the following:

  • "jitter" (default): plots a scatter plot with jitter per independent variable and adds a linear regression line with 95% confidence intervals to it; keep in mind that if you have, say, three independent variables, this visualization shows you three plots with one linear regression for each, so that the three models (i.e., the three colored lines) reflect only the particular combination of one independent and the dependent variable

  • "alpha" (default): almost like jitter but instead of jitter it plots scatter plots with some transparency so that multiple data points in the same position appear as darker

  • "correlogram": like to_correlation_matrix(), a correlogram between independent variables are produced to help determine independent errors and multicollinearity

  • "residualsfitted" or "resfit": a residuals-versus-fitted plot is useful to determine distributions; for a normal distribution the colored line should ideally fit on the dashed line

  • "pp": a (normal) probability-probability plot helps checking for multicollinearity whereby the data (here mostly the center data from within the IQR) should ideally align with the dashed line

  • "qq": a (normal) quantile-quantile plot helps checking for multicollinearity but focuses more on outliers; the data should align with the dashed line

  • "scalelocation" or "scaloc": a scale-location (sometimes also called a spread-location) plot checks whether residuals are spread equally to help check for homoscedasticity; ideally, the colored line is horizontal and the data spreads more or less randomly

  • "residualsleverage" or "reslev": a residuals-versus-leverage plot allows to check for influential outliers affecting the final model more than the rest of the data; ideally, no data is far off compared to the bulk of the the data and thus shows high Cook's distance to the rest; the colored line helps to identify the bulk of the data and the five most-distant outliers are labelled with their case number (i.e., the row number in the dataset); note that 5 is arbitrary here, meaning that they might not be too far off or there might be more than 5 noteworthy outliers in this model; interpret with care

Note that the returned ggplot2 object can be modified easily by appending or overwriting individual geom's or scale's. See the examples below and the documentation of ggplot2.

Value

A ggplot2 object

Examples

## Not run: 
WoJ %>%
  describe() %>%
  visualize()

fbposts %>%
  describe_cat() %>%
  visualize()

WoJ %>%
  tab_frequencies(trust_parliament) %>%
  visualize()
fbposts %>%
  tab_frequencies(pop_elite, pop_people, pop_othering) %>%
  visualize()

WoJ %>%
  crosstab(reach, employment) %>%
  visualize()

fbposts %>%
  crosstab(coder_id, type, percentages = TRUE) %>%
  visualize()

WoJ %>%
  t_test(temp_contract, autonomy_selection, autonomy_emphasis) %>%
  visualize()

WoJ %>%
  unianova(country, autonomy_selection, autonomy_emphasis) %>%
  visualize()

fbposts %>%
  correlate(pop_elite, pop_people) %>%
  visualize()

fbposts %>%
  correlate(pop_elite, pop_people, with = pop_othering) %>%
  visualize()

fbposts %>%
  correlate(pop_elite, pop_people) %>%
  visualize("alpha")

WoJ %>%
  correlate(autonomy_selection, ethics_1, partial = work_experience) %>%
  visualize()

WoJ %>%
  correlate(ethics_1, ethics_2, ethics_3, ethics_4) %>%
  to_correlation_matrix() %>%
  visualize()

r <- WoJ %>% regress(autonomy_selection, temp_contract, work_experience, ethics_2)
r %>% visualize() # same as r %>% visualize("jitter")
r %>% visualize("alpha")
r %>% visualize("correlogram")
r %>% visualize("resfit")
r %>% visualize("pp")
r %>% visualize("qq")
r %>% visualize("scaloc")
r %>% visualize("reslev")

# To overwrite a certain scale or geom, just append as you would with ggplot2
fbposts %>%
  describe_cat() %>%
  visualize() +
    ggplot2::scale_fill_grey()

## End(Not run)

Z-standardize numeric, continuous variables

Description

This function z-standardizes the specified numeric columns or all numeric columns if none are specified. A z-standardized scale centers at a mean of 0.0 and has a standard deviation of 1.0, making it comparable to other z-standardized distributions.

Usage

z_scale(data, ..., name = NULL, overwrite = FALSE)

Arguments

Value

A tdcmm model with the z-standardized variable(s).

See Also

Other scaling: categorize_scale(), center_scale(), dummify_scale(), minmax_scale(), recode_cat_scale(), reverse_scale(), setna_scale()

Examples

WoJ %>% z_scale(autonomy_emphasis)
WoJ %>% z_scale(autonomy_emphasis, name = "my_zstdized_variable")
WoJ %>%
  z_scale(autonomy_emphasis) %>%
  tab_frequencies(autonomy_emphasis, autonomy_emphasis_z)

Determine if range of vector is close to zero, with a specified tolerance

Description

The machine epsilon is the difference between 1.0 and the next number that can be represented by the machine. By default, this function uses epsilon * 1000 as the tolerance. First it scales the values so that they have a mean of 1, and then it checks if the difference between them is larger than the tolerance.

Usage

zero_range(x, tol = 1000 * .Machine$double.eps)

Arguments

Value

logical TRUE if the relative difference of the endpoints of the range are not distinguishable from 0.