splithalfr: Split-Half Reliabilities

Description

Estimates split-half reliabilities for scoring algorithms of cognitive tasks and questionnaires.

Getting started

We've got six short vignettes to help you get started. You can open a vignette by running the corresponding code snippet (vignette(...)) in the R console.

• vignette("rapi_sum") Sum-score for data of the 23-item version of the Rutgers Alcohol Problem Index (White & Labouvie, 1989 <doi:10.15288/jsa.1989.50.30>)

• vignette("vpt_diff_of_means") Difference of mean RTs for correct responses, after removing RTs below 200 ms and above 520 ms, on Visual Probe Task data (Mogg & Bradley, 1999 <doi:10.1080/026999399379050>)

• vignette("aat_double_diff_of_medians") Double difference of medians for correct responses on Approach Avoidance Task data (Heuer, Rinck, & Becker, 2007 <doi:10.1016/j.brat.2007.08.010>)

• vignette("iat_dscore_ri") Improved d-score algorithm for data of an Implicit Association Task that requires a correct response in order to continue to the next trial (Greenwald, Nosek, & Banaji, 2003)

• vignette("sst_ssrti") Stop-Signal Reaction Time integration method for data of a Stop Signal Task (Logan, 1981)

• vignette("gng_dprime") D-prime for data of a Go/No Go task (Miller, 1996 <doi:10.3758/BF03205476>)

Splitting methods

The splithalfr supports a variety of methods for splitting your data. We review and assess each method in the compendium paper (Pronk et al., 2021 <doi:10.3758/s13423-021-01948-3>), but based on more recent concerning findings with Monte Carlo splitting (Kahveci et al., 2025 <doi:10.3758/s13423-024-02597-y>; Pronk et al., 2023 <doi:10.3758/s13428-022-01885-6>), I now only recommend permutated splitting and not Monte Carlo splitting. This vignette illustrates how to apply each splitting method via the splithalfr: 'vignette("splitting_methods")'

• first-second and odd-even (Green et al., 2016 <doi:10.3758/s13423-015-0968-3>; Webb, Shavelson, & Haertel, 1996 <doi:10.1016/S0169-7161(06)26004-8>; Williams & Kaufmann, 2012 <doi:10.1016/j.jesp.2012.03.001>)

• stratified (Green et al., 2016 <doi:10.3758/s13423-015-0968-3>)

• permutated/bootstrapped/random sample of split halves (Kopp, Lange, & Steinke, 2021 <doi:10.1177/1073191119866257>, Parsons, Kruijt, & Fox, 2019 <doi:10.1177/2515245919879695>; Williams & Kaufmann, 2012 <doi:10.1016/j.jesp.2012.03.001>)

• Monte Carlo (Williams & Kaufmann, 2012 <doi:10.1016/j.jesp.2012.03.001>)

Citing the splithalfr

Please cite the compendium paper (Pronk et al., 2022 <doi:10.3758/s13423-021-01948-3>]) and the software. To cite the software, type 'citation("splithalfr")' in R, or use the reference below.

Pronk, T. (2025). splithalfr: Estimates split-half reliabilities for scoring algorithms of cognitive tasks and questionnaires (Version 3.0.0) [Computer software]. <doi:10.5281/zenodo.7777894>

Validation of split-half estimations

Part of the splithalfr algorithm has been validated via a set of simulations that are not included in this package. The R script for these simulations can be found here.

Related packages

These R packages offer bootstrapped split-half reliabilities for specific scoring algorithms and are available via CRAN at the time of this writing: multicon, psych, splithalf, and rapidsplithalf.

Acknowledgments

I would like to thank Craig Hedge, Eva Schmitz, Fadie Hanna, Helle Larsen, Marilisa Boffo, and Marjolein Zee for making datasets available for inclusion in the splithalfr. Additionally, I would like to thank Craig Hedge and Benedict Williams for sharing R-scripts with scoring algorithms that were adapted for splithalfr vignettes. Finally, I would like to thank Mae Nuijs and Sera-Maren Wiechert for spotting bugs in earlier versions of this package.

Author(s)

Maintainer: Thomas Pronk pronkthomas@gmail.com

See Also

Useful links:

• https://github.com/tpronk/splithalfr

• Report bugs at https://github.com/tpronk/splithalfr/issues

Calculate Angoff-Feldt coefficient

Description

Angoff-Feldt reliability coefficient. Formula obtained from Warrens (2015) <doi:10.1007/s11634-015-0198-6>

Usage

angoff_feldt(x, y)

Arguments

Value

(numeric) Angoff-Feldt cefficient

See Also

Other splithalfr coefficients: assmd(), flanagan_rulon(), sdregi(), short_icc(), spearman_brown()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate coefficient
angoff_feldt(vars[,1], vars[,2])

Split each element of a list of strata based on a list of indexes

Description

Splits each element of strata into two parts based on a list of indexes. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

apply_split_indexes_to_strata(strata, indexes)

Arguments

Value

(list) A list with two elements, containing the first and second split of strata.

See Also

Other splitting functions: apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), split_stratum(), stratify()

Examples

# Stratify a data frame, then split it odd-even
ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
strata <- stratify(ds, ds$condition)
split_indexes <- get_split_indexes_from_strata(strata, method = "odd_even")
apply_split_indexes_to_strata(strata, split_indexes)

Split a stratum based on a list of indexes

Description

Splits stratum into two parts based on a list of indexes. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

apply_split_indexes_to_stratum(stratum, indexes_1, indexes_2)

Arguments

Value

(list) List with two elements that contain stratum split in two parts.

See Also

Other splitting functions: apply_split_indexes_to_strata(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), split_stratum(), stratify()

Examples

# Random split-half. One of the splits gets 4 elements and the other 5
stratum =  letters[1:9]
indexes = get_split_indexes_from_stratum(stratum)
apply_split_indexes_to_stratum(stratum, indexes[[1]], indexes[[2]])

Calculate Absolute Strictly Standardized Mean Difference (ASSMD)

Description

Returns the absolute difference of the mean of x and y divided by their shared standard deviation. Since the resulting difference is absolute, the larger of the two means is always used as minuend and the smallest as subtrahend. Based on Zhang (2012) <doi:10.1016/j.ygeno.2006.12.014>

Usage

assmd(x, y)

Arguments

Value

(numeric) Absolute SSMD

See Also

Other splithalfr coefficients: angoff_feldt(), flanagan_rulon(), sdregi(), short_icc(), spearman_brown()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate Absolute SSMD
assmd(vars[,1], vars[,2])

Calculate split scores per participant

Description

Calculates split scores, by applying fn_score to subsets of data as specified via participants. It provides a range of additional arguments for different splitting methods and to support parallel processing. To learn more about writing scoring algorithms for use with the splithalfr, see the included vignettes. by_split is modeled after the by function, accepting similar values for the first three arguments (data, INDICES, FUN). For more information about different methods for splitting data, see get_split_indexes_from_stratum. For more information about stratification, see split_df

Usage

by_split(
  data,
  participants,
  fn_score,
  stratification = NULL,
  replications = 1,
  method = c("random", "odd_even", "first_second"),
  replace = FALSE,
  split_p = 0.5,
  subsample_p = 1,
  subsample_n = NULL,
  careful = TRUE,
  match_participants = FALSE,
  ncores = detectCores(),
  seed = NULL,
  verbose = TRUE
)

Arguments

Value

(data frame) Returns a data frame with a column for participant, a column replication that counts split replications, and score_1 and score_2 for the score calculated of each part via fn_score.

Examples

# N.B. This example uses R script from the vignette: "rapi_sum"
data("ds_rapi", package = "splithalfr")
# Convert to long format
ds_long <- reshape(
  ds_rapi,
  varying = paste("V", 1 : 23, sep = ""),
  v.names = "answer",
  direction = "long",
  idvar = "twnr",
  timevar = "item"
)
# Function for RAPI sum score
rapi_fn_score <- function (data) {
  return (sum(data$answer))
}
# Calculate scores on full data
by(
  ds_long,
  ds_long$twnr,
  rapi_fn_score
)
# Permutation split, one iteration, items matched across participants
split_scores <- by_split(
  ds_long,
  ds_long$twnr,
  rapi_fn_score,
  ncores = 1,
  match_participants = TRUE
)
# Mean flanagan-rulon coefficient across splits
fr <- mean(split_coefs(split_scores, flanagan_rulon))

Check whether two strata have the same structure

Description

Checks strata against strata_left. Each element of strata_left should also be present in strata, be of a similar type (data frame/tibble or list/vector), and be of similar size (nrow for data frames/tibbles or length for lists/vectors). Stops with an arror if any checks fail.

Usage

check_strata(strata_left, strata_right)

Arguments

Value

None

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), split_stratum(), stratify()

Examples

check_strata(list(1 : 4), list(1 : 4))

Example Approach Avoidance Task (AAT) Measurement Data in JASMIN2 Format

Description

The JASMIN1 AAT was an irrelevant feature task, in which participants were instructed to approach/avoid left/right rotated stimuli. This particular AAT was administered (and described in detail) in Boffo et al. (2018) <doi:10.1111/add.14071>. Participants were presented stimuli from a "test" category, which were gambling-related pictures, and from a "control" category, which were pictures unrelated to gambling. It registered approach responses by participants pressing (and holding) the arrow down key, while avoid responses were given via the arrow up key. Upon a response, the stimulus zoomed in or out, until it disappeared from the screen. The first response to a stimulus was logged. The dataset contains one row per trial. This dataset was graciously provided by Eva Schmitz.

Usage

ds_aat

Format

An object of class data.frame with 6528 rows and 12 columns.

Details

Overview of columns:

• UserID. Identifies participants

• approach_tilt. If "left", participants were instructed to approach left rotated stimuli. If "right", participants were instructed to approach right rotated stimuli.

• block_type. Type of block: "practice" for practice trials with neutral stimuli, "assess" for assessment trials with salient stimuli

• block. Counts blocks, starting at zero

• trial. Counts trials in blocks, starting at zero

• appr. If "yes", this trial was an approach trial. If "no", this trial was an avoid trial.

• tilt. Whether the stimulus was rotated to the "left" or to the "right"

• cat. Stimulus category: "practice", "test", or "control"

• stim. Stimulus ID

• response. Response; 1 = correct, 2 = incorrect, 3 = timeout (no response in 4000 ms), 4 = invalid key

• rt. Response time in milliseconds

• sust. Was approach or avoid response sustained until the stimulus was completely zoomed in or out?

Example Go/No Go data

Description

The Go/No Go is a task in which participants respond to one set of stimuli, but withhold a response to another set of stimuli. This particular dataset is from the first session of a study that is described in detail in Hedge et al. (2018) <doi:10.1016/j.tate.2019.102887>. It was graciously provided by Craig Hedge and can be obtained from https://osf.io/cwzds.

Usage

ds_gng

Format

An object of class data.frame with 28200 rows and 7 columns.

Details

Overview of columns:

• block. Block number

• trial. Trial number

• stim. Stimuli set used in that block

• condition. 0 = go, 2 = no go

• response. Correct (1) or incorrect (0)

• rt. Reaction time (seconds)

• participant. Participant ID

Example Implicit Association Task (IAT) Data in JASMIN2 Format

Description

The JASMIN2 IAT closely followed the original IAT procedure (Greenwald, McGhee, & Schwartz, 1998), except that target and attribute trials did not alternate. Upon a correct response, the next trial started. Upon an incorrect response, the current trial was repeated. The response to each trial was logged. This particular dataset is from a Ethnicity-Valence IAT, which was administered (and described in detail) in Abacioglu et al. (2019) <doi:10.1016/j.tate.2019.102887>. This dataset was graciously provided by Fadie Hanna and Marjolein Zee.

Usage

ds_iat

Format

An object of class data.frame with 9696 rows and 11 columns.

Details

Overview of columns:

• participation_id Identifies participants

• t1_left. If TRUE, the first combination block had target 1 on the left (and target 2 on the right)

• a1_left. If TRUE, the first combination block had attribute 1 on the left (and attribute 2 on the right)

• block_type. Type of block

• block. Counts blocks, starting at zero

• trial. Counts trials in blocks, starting at zero

• attempt. Counts attempts (responses) in trials, starting at zero

• cat. Category that stimulus belonged to

• stim. Stimulus

• response. Response; 1 = correct, 2 = incorrect, 3 = timeout (no response in 4000 ms), 4 = invalid key

• rt. Response time in milliseconds. Note that some response times may exceed the timeout window due to clock errors on the computer that the IAT was administered

The variable block_type can have these values:

• tar_discr: target discrimination

• att_discr: attribute discrimination

• tar1att1_1: target 1 with attitude 1, practice block

• tar1att1_2: target 1 with attitude 1, test block

• tar_rev: reverse target discrimination

• tar1att2_1: target 1 with attitude 2, practice block

• tar1att2_2: target 1 with attitude 2, test block

Example 23-item Rutgers Alcohol Problem Inventory (RAPI) data

Description

The RAPI is a questionnaire which asks how often a participant experienced each of 23 alcohol-related problems within the last year (White & Labouvie, 1989 <doi:10.15288/jsa.1989.50.30>). The dataset contains one row per participant.

Usage

ds_rapi

Format

An object of class data.frame with 426 rows and 24 columns.

Details

The dataset contains the following columns:

• twnr. Identifies participants

• V1 to V23. Answers on each of the 23 RAPI items

Each item is answered on a four-point scale with the following answer options:

• 0 = None

• 1 = 1-2 times

• 2 = 3-5

• 3 = More than 5 times

Example Stop Signal Task data

Description

The Stop Signal Task is a task in which participants responded whether a stimulus was a square or a circle. On 25 This particular dataset is from the first session of a study that is described in detail in Hedge et al. (2018) <doi:10.1016/j.tate.2019.102887>. It was graciously provided by Craig Hedge and can be obtained from https://osf.io/cwzds.

Usage

ds_sst

Format

An object of class data.frame with 27000 rows and 7 columns.

Details

Overview of columns:

• block. Block number

• trial. Trial number

• ssd. Stop signal delay

• condition. 0 = go, 1 = stop

• response. Correct (1) or incorrect (0)

• rt. Reaction time (milliseconds)

• participant. Participant ID

Example Visual Probe Task (VPT) Measurement Data in JASMIN1 Format

Description

The JASMIN1 VPT distinguished between "test" stimuli, which are in some way assumed to be salient to the participant and "control" stimuli, which are not. Test and control stimuli were presented in pairs, with one left and one right, followed by a probe that was an arrow pointing up or down. Participants needed to indicate whether the arrow pointed up or down. Upon a correct response the next trial started and upon an incorrect response the current trial was repeated. Only the first response to a new trial was logged. This particular VPT was part of the pre-measurement of a cognitive bias modification study. The "test" stimuli were alcoholic beverages and the "control" stimuli were non-alcoholic beverages, selected from the Amsterdam Beverage Picture Set (Pronk et al., 2015) <doi:10.1111/acer.12853>. The dataset contains one row per trial. This dataset was graciously provided by Marilisa Boffo.

Usage

ds_vpt

Format

An object of class data.frame with 19520 rows and 12 columns.

Details

Overview of columns:

• UserID. Identifies participants

• patt. Probe-at-test. If "yes", the probe was positioned at the test stimulus. If "no", the probe was positioned at the control stimulus.

• phor. Probe horizontal position. Values: "left" or "right"

• thor. Test horizontal position. Values: "left" or "right"

• keep. If "yes" the probe was superimposed on the stimuli. If "no" the probe replaced the stimuli.

• pdir. Probe direction. Values: "up" or "down"

• stim. Stimulus

• response. Response; 1 = correct, 2 = incorrect, 3 and NA = timeout (no response in 5000 ms), 4 = invalid key

• rt. Response time in milliseconds

• block. Counts blocks, starting at zero

• trial. Counts trials in blocks, starting at zero

• block_type. Type of block: "assess" for assessment trials with salient stimuli

Calculate Flanagon-Rulon coefficient

Description

Flanagon-Rulon reliability coefficient. Formula obtained from Warrens (2015) <doi:10.1007/s11634-015-0198-6>

Usage

flanagan_rulon(x, y)

Arguments

Value

(numeric) Flanagon-Rulon coefficient

See Also

Other splithalfr coefficients: angoff_feldt(), assmd(), sdregi(), short_icc(), spearman_brown()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate coefficient
flanagan_rulon(vars[,1], vars[,2])

Generate indexes for splitting strata

Description

Generates indexes for splitting each element of strata into two parts. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

get_split_indexes_from_strata(strata, ...)

Arguments

Value

(list) A list with two elements, containing the first and second part of strata.

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), split_stratum(), stratify()

Examples

# Stratify a data frame, then split it odd-even
ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
strata <- stratify(ds, ds$condition)
split_indexes <- get_split_indexes_from_strata(strata, method = "odd_even")
apply_split_indexes_to_strata(strata, split_indexes)

Generate indexes that can be used to split a stratum into two parts

Description

get_split_indexes_from_stratum returns a list with indexes for splitting its stratum argument in two parts. The splits differ at most by one in size. With default arguments, a random split-half is returned, which samples elements for each part from stratum without replacement. Via additional arguments to get_split_indexes_from_stratum a range of other splitting methods can be applied.

Usage

get_split_indexes_from_stratum(
  stratum,
  method = c("random", "odd_even", "first_second"),
  replace = FALSE,
  split_p = 0.5,
  subsample_p = 1,
  careful = TRUE
)

Arguments

Details

The following rounding rules apply to subsample size and split size:

• If the size of the subsample, calculated as subsample_p times size of stratum, is a fraction, then subsample size is rounded up.

• If the joint size of the two parts, calculated as 2 * split_p times size of the subsampled stratum, is a fraction, the part size is rounded up.

• If the joint size of the two parts is odd and replace is FALSE, then one of the parts randomly gets one more element than the other part.

• If the joint size of the two parts is odd and replace is TRUE, part size is rounded up to the next whole number, so each of the splits has the same size.

Value

(list) List with two elements that contain indexes that can be used to split the stratum in two parts two splits of stratum.

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), split_df(), split_strata(), split_stratum(), stratify()

Examples

# Split-half. One of the splits gets 4 elements and the other 5
stratum = letters[1:9]
indexes = get_split_indexes_from_stratum(stratum)
apply_split_indexes_to_stratum(stratum, indexes[[1]], indexes[[2]])

SD ratio of equalities or greater inequalities

Description

Returns the ratio of the SDs of x and y, using the largest SD of the two as denominator. Hence, the result is always 1 (ratio of equalities) or greater than 1 (ratio of greater inequalities). If x or y have less than two elements, NA is returned.

Usage

sdregi(x, y)

Arguments

Value

(numeric) SD ratio

See Also

Other splithalfr coefficients: angoff_feldt(), assmd(), flanagan_rulon(), short_icc(), spearman_brown()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate SD ratio of left and right variables
sdregi(vars[,1], vars[,2])
# Calculate SD ratio of right and left variables; should give same result
sdregi(vars[,1], vars[,2])

Calculate Intraclass Correlation Coefficient (ICC)

Description

Wrapper for ICCs calculated via ICC. If x or y have less than two elements, NA is returned.

Usage

short_icc(
  x,
  y,
  type = c("ICC1", "ICC2", "ICC3", "ICC1k", "ICC2k", "ICC3k"),
  ...
)

Arguments

Value

(numeric) Value of ICC coefficient

See Also

Other splithalfr coefficients: angoff_feldt(), assmd(), flanagan_rulon(), sdregi(), spearman_brown()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate ICC1
short_icc(vars[,1], vars[,2], type = "ICC1", lmer = FALSE)

Calculate Spearman-brown coefficient

Description

Spearman-Brown reliability coefficient for doubling test length. Formula obtained from Warrens (2015) <doi:10.1007/s11634-015-0198-6>

Usage

spearman_brown(x, y, fn_cor = cor, ...)

Arguments

Value

(numeric) Spearman-Brown coefficient

See Also

Other splithalfr coefficients: angoff_feldt(), assmd(), flanagan_rulon(), sdregi(), short_icc()

Examples

# Generate two variables with different means, variances and a correlation of about 0.5
library(MASS)
vars = mvrnorm(30, mu = c(0, 2), Sigma = matrix(c(5, 2, 2, 3), ncol = 2), empirical = TRUE)
# Calculate coefficient based on Pearson correlation
spearman_brown(vars[,1], vars[,2])
# Calculate coefficient based on ICC, two-way, random effects, absolute agreement, single rater
spearman_brown(vars[,1], vars[,2], short_icc, type = "ICC1", lmer = FALSE)

Generate bootstrap replicates of a coefficient split-half reliability estimate by sampling participants

Description

Generates bootstrap replicates via boot. The parameter ds should be a data frame as returned by by_split: Each unique value of the column participant is considered a independent sample of the target population. For each unique value of the column split in ds, it selects the corresponding rows in ds, and passes the values in the columns score_1 and score_2 as the first and second argument to fn_coef. Any row in ds for which score_1 or score_2 is NA is pairwise removed before passing the data to fn_coef. Any coefficient that is NA is removed before passing the data to fn_summary.

Usage

split_ci(
  ds,
  fn_coef,
  fn_average,
  bootstrap_replications = 1000,
  parallel = "snow",
  ncpus = detectCores(),
  ...
)

Arguments

Details

For a practical example, see one of the vignettes for getting started with the splithalfr. Also, note that the boot package supports parallel processing via the parameters 'parallel' and 'ncpus'.

For averaging internal consistency coefficients, see Feldt and Charter (2006). For more information about bias-corrected and accelerated bootstrap confidence intervals, see Efron (1987).

Value

Confidence interval

References

Efron, B. (1987). Better bootstrap confidence intervals. Journal of the American statistical Association, 82(397), 171-185. doi:10.1080/01621459.1987.10478410

Feldt, L. S., & Charter, R. A. (2006). Averaging internal consistency reliability coefficients. Educational and Psychological Measurement, 66(2), 215-227. doi:10.1177/0013164404273947

See Also

Other split aggregation functions: split_coefs()

Examples

# Import boot library
library(boot)
# Generate five splits with scores that are correlated 0.00, 0.25, 0.5, 0.75, and 1.00
library(MASS)
ds_splits = data.frame(V1 = numeric(), V2 = numeric(), split = numeric())
for (r in 0:4) {
  vars = mvrnorm(10, mu = c(0, 0), Sigma = matrix(c(10, 3, 3, 2), ncol = 2), empirical = FALSE)
  ds_splits = rbind(ds_splits, cbind(vars, r, 1 : 10))
}
names(ds_splits) = c("score_1", "score_2", "replication", "participant")
# Conduct bootstrap
bootstrap_result <- split_ci(ds_splits, cor, mean, parallel = "no")
# Get boosted and accelerated confidence intervals
print(boot.ci(bootstrap_result, type="bca"))

Calculate a bivariate coefficient for each split-half replication

Description

Calculates a bivariate coefficient across participants for each split-half replication and returns their values calculated across replications. ds should be a data frame as returned by by_split: For each unique value of the column split in ds, it selects the corresponding rows in ds, and passes the values in the columns score_1 and score_2 as the first and second argument to fn_coef. Any row in ds for which score_1 or score_2 is NA is pairwise removed before passing the data to fn_coef. For averaging internal consistency coefficients, see Feldt and Charter (2006).

Usage

split_coefs(ds, fn_coef, ...)

Arguments

Value

Coefficients per split calculated via fn_coef.

References

Feldt, L. S., & Charter, R. A. (2006). Averaging internal consistency reliability coefficients. Educational and Psychological Measurement, 66(2), 215-227. doi:10.1177/0013164404273947

See Also

Other split aggregation functions: split_ci()

Examples

# Generate five splits with scores that are correlated 0.00, 0.25, 0.5, 0.75, and 1.00
library(MASS)
ds_splits = data.frame(score_1 = numeric(), score_2 = numeric(), replication = numeric())
for (r in 0:4) {
  vars = mvrnorm(10, mu = c(0, 0), Sigma = matrix(c(10, 3, 3, 2), ncol = 2), empirical = FALSE)
  ds_splits = rbind(ds_splits, cbind(vars, r))
}
names(ds_splits) = c("score_1", "score_2", "replication")
# Pearson correlations
split_coefs(ds_splits, cor)
# Spearman-brown corrected Pearson correlations
split_coefs(ds_splits, spearman_brown)
# Flanagan-Rulon coefficient
split_coefs(ds_splits, flanagan_rulon)
# Angoff-Feldt coefficient
split_coefs(ds_splits, angoff_feldt)
# Spearman-Brown corrected ICCs
split_coefs(
  ds_splits,
  spearman_brown,
  short_icc,
  type = "ICC1",
  lmer = FALSE
)

Split a data frame into two parts

Description

Splits data, Applies a stratified split to a data frame and returns each part. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

split_df(data, stratification = NULL, ...)

Arguments

Value

(list) List with two elements that each contain one of two parts.

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_strata(), split_stratum(), stratify()

Examples

ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
split_df(ds, method = "random")
split_df(ds, method = "odd_even")
split_df(ds, method = "first_second")
split_df(ds, stratification = ds$condition, method = "random")
split_df(ds, stratification = ds$condition, method = "odd_even")
split_df(ds, stratification = ds$condition, method = "first_second")
ds <- data.frame(condition = rep(c("a", "b"), 4), score = 1 : 8)
split_df(ds, method = "random")
split_df(ds, method = "odd_even")
split_df(ds, method = "first_second")
split_df(ds, stratification = ds$condition, method = "random")
split_df(ds, stratification = ds$condition, method = "odd_even")
split_df(ds, stratification = ds$condition, method = "first_second")

Split each stratum into two parts

Description

Splits each element of strata into two parts. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

split_strata(strata, ...)

Arguments

Value

(list) A list with two elements, containing the first and second split of strata.

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_stratum(), stratify()

Examples

# Stratify a data frame, then split it odd-even
ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
strata <- stratify(ds, ds$condition)
split_strata(strata, method = "odd_even")

Split a stratum into two parts

Description

Splits stratum into two parts. For more information about splitting options, and an extensive list of examples, see get_split_indexes_from_stratum.

Usage

split_stratum(stratum, ...)

Arguments

Value

(list) List with two elements that contain each of the two parts of stratum split in two.

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), stratify()

Examples

# Split stratum odd-even
ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
split_stratum(ds, method = "odd_even")

Stratify a data frame

Description

Split a dataframe into strata formed by each a unique value of stratification.

Usage

stratify(ds, stratification = NULL)

Arguments

Value

(list) List of strata

See Also

Other splitting functions: apply_split_indexes_to_strata(), apply_split_indexes_to_stratum(), check_strata(), get_split_indexes_from_strata(), get_split_indexes_from_stratum(), split_df(), split_strata(), split_stratum()

Examples

# Stratify a data frame, then split it odd-even
ds <- data.frame(condition = rep(c("a", "b"), each = 4), score = 1 : 8)
strata <- stratify(ds, ds$condition)
split_strata(strata, method = "odd_even")