Package 'DACF'

cf_correction

Description

Extends Tukey's HSD and Bonferroni procedure to account for ceiling and floor effects

Usage

cf_correction(
  x,
  tests = "all",
  df.adjustment = "trunc",
  gh.correction = "no_gh",
  alpha = 0.05,
  flr,
  ceil
)

Arguments

x	a dataframe of data with ceiling/floor effects and corresponding group variables or aov results
tests	a character string specifying the desired multiple-comparison procedure: "all" (default, Tukey's HSD and Bonferroni), "tukey" (Tukey's HSD), or "bonf" (Bonferroni)
df.adjustment	a character string specifying the desired method for adjusting the degree of freedom: "trunc" (default, Liu and Wang's truncated-normal corrections) or "unadj" (unadjusted)
gh.correction	a character string specifying if the Welch & Games-Howell correction for heteroscedasticity should be included: "no_gh"(default) or "yes_gh"
alpha	a (non-empty) numeric value of desired of alpha level
flr	a (non-empty) numeric value of the floor threshold (e.g., the minimum score of the measurement scale)
ceil	a (non-empty) numeric value of the ceiling threshold (e.g., the maximum score of the measurement scale)

Value

a matrix containing pairwise comparison results. Columns depend on the tests and gh.correction arguments:

• Comparison_i: First group in the comparison

• Comparison_j: Second group in the comparison

• mean_i: The adjusted mean of the first group

• mean_j: The adjusted mean of the second group

• diff_in_means: The difference in adjusted means

• tukey.CI_lwr: Lower bound of Tukey's HSD confidence interval (tests = "tukey" or "all")

• tukey.CI_upr: Upper bound of Tukey's HSD confidence interval (tests = "tukey" or "all")

• hedges_g: Hedges' g effect size (tests = "tukey" or "all")

• Q: Adjusted Tukey's Q statistic (tests = "tukey" or "all")

• p: p-value for Tukey's Q statistic (tests = "tukey" or "all")

• t: t-test statistic (tests = "bonf" or "all")

• t_p: p-value for the t-test statistic (tests = "bonf" or "all")

• p.bonferroni: Bonferroni-adjusted p-value (tests = "bonf" or "all")

• bonf.CI_lwr: Lower bound of Bonferroni confidence interval (tests = "bonf" or "all")

• bonf.CI_upr: Upper bound of Bonferroni confidence interval (tests = "bonf" or "all")

• gh.CI_lwr: Lower bound of Games-Howell confidence interval (gh.correction = "yes_gh", tests = "tukey" or "all")

• gh.CI_upr: Upper bound of Games-Howell confidence interval (gh.correction = "yes_gh", tests = "tukey" or "all")

• games.howell.t: Games-Howell t statistic (gh.correction = "yes_gh", tests = "tukey" or "all")

• games.howell.p: Games-Howell p-value (gh.correction = "yes_gh", tests = "tukey" or "all")

• games.howell.g: Games-Howell Hedges' g (gh.correction = "yes_gh", tests = "tukey" or "all")

• welch.satterthwaite.df: Welch-Satterthwaite degrees of freedom (gh.correction = "yes_gh")

• welch.bonf.CI_lwr: Lower bound of Welch-Bonferroni confidence interval (gh.correction = "yes_gh", tests = "bonf" or "all")

• welch.bonf.CI_upr: Upper bound of Welch-Bonferroni confidence interval (gh.correction = "yes_gh", tests = "bonf" or "all")

• t_Welch: Welch t statistic (gh.correction = "yes_gh", tests = "bonf" or "all")

• p_Welch: Welch p-value (gh.correction = "yes_gh", tests = "bonf" or "all")

• p_Welch.bonferroni: Welch Bonferroni-adjusted p-value (gh.correction = "yes_gh", tests = "bonf" or "all")

References

Aitkin MA. Correlation in a Singly Truncated Bivariate Normal Distribution. Psychometrika. 1964;29(3):263-270. doi:10.1007/BF02289723

Cohen, A. C. (1959). Simplified Estimators for the Normal Distribution When Samples Are Singly Censored or Truncated. AnnalsTechnometrics, 1(3), 217–237. doi:10.1080/00401706.1959.10489859

Greene, W. H. (2002). Econometric Analysis. In Econometric Analysis.

Liu, Q., Wang, L. (2020) t-Test and ANOVA for data with ceiling and/or floor effects. Behav Res 53, 264–277 . doi:10.3758/s13428-020-01407-2

Qi, H., Dong, Z., Wei, Q., Chen, X., & Luo, Y. (2025). Are Gamers Happier? Multidimensional Well-Being Differences in Risk Groups for Problematic Internet Use and Internet Gaming Disorder. Personality and Individual Differences 246, 113334. doi:10.2139/ssrn.5148047

Examples

data("Qi_data_males") # (Qi et al., 2025)
cf_correction(Qi_data_males, tests = "all", df.adjustment = "trunc",
              gh.correction = "yes_gh", alpha = .05, flr = 0, ceil = 9)
cf_correction(Qi_data_males, tests = "tukey", df.adjustment = "unadj",
              gh.correction = "no_gh", alpha = .05, flr = 0, ceil = 9)
cf_correction(Qi_data_males, tests = "all", df.adjustment = "unadj",
              gh.correction = "no_gh", alpha = .05, flr = 0, ceil = 9)

---

f.star.test

Description

conduct a Brown-Forsythe F star test

Usage

f.star.test(means, variances, ns)

Arguments

means	a (non-empty) numeric vector of the group means
variances	a (non-empty) numeric vector of the group variances
ns	a (non-empty) numeric vector of sample sizes per group

Value

statistic	the value of the adjusted Brown-Forsythe F star statistic
p.value	the p-value for the test
est.f.squared	effect size estimate as in Cohen's f squared

Examples

# a f star test for three-group mean comparison
f.star.test(c(-.2,0,.2),c(1,1,1),c(100,100,100))
f.star.test(c(0,0,1),c(2,1,3),c(100,100,100))

---

Gaussian Graphical Model with Ceiling/Floor Effect Correction

Description

Estimates a psychological network (GGM) from data with ceiling and/or floor effects using a two-step pairwise censored-normal correction. Step 1 corrects item means and variances via truncated-normal moment matching (Liu & Wang, 2021). Step 2 estimates pairwise latent correlations by solving the moment equation E[cov*(rho)] = observed censored covariance, evaluated using 50-point Gauss-Hermite quadrature. The corrected covariance matrix is supplied as input to EBICglasso or significance-based edge selection, preserving the standard qgraph/bootnet workflow.

Usage

ggm_cfe(
  data,
  floor = NULL,
  ceiling = NULL,
  method = "EBICglasso",
  gamma = 0.5,
  nlambda = 100,
  lambda.min.ratio = 0.01,
  threshold = FALSE,
  alpha = 0.05,
  verbose = TRUE
)

Arguments

data	A data frame or numeric matrix (n x p). Rows = observations, columns = variables (items/nodes).
floor	Floor threshold(s). Either a single value applied to all variables, or a named numeric vector with one value per column. Use NULL to indicate no floor effect. Default: NULL.
ceiling	Ceiling threshold(s). Same format as floor. Default: NULL.
method	Estimation method(s): "EBICglasso" (default), "FDR", "Bonferroni", or "all" to run all three.
gamma	EBIC hyperparameter (0 = BIC, 0.5 = default). Only used when method includes "EBICglasso".
nlambda	Number of tuning parameters searched by EBICglasso. Default: 100.
lambda.min.ratio	Smallest lambda searched, as a fraction of the largest lambda. Passed to qgraph::EBICglasso. Default: 0.01.
threshold	Logical. Passed to qgraph::EBICglasso; TRUE enforces higher specificity at the cost of sensitivity. Default: FALSE.
alpha	Significance level for p-value methods. Default: 0.05.
verbose	Logical. Print progress and diagnostics. Default: TRUE.

Value

An object of class "ggm_cfe" containing:

network	Partial correlation matrix from the corrected method (primary estimator: EBICglasso or first specified method).
network_corrected	Partial correlation matrix from naive EBICglasso (ignoring ceiling/floor effects).
network_naive	Partial correlation matrix estimated from the raw (uncorrected) sample covariance using the primary method.
networks	Named list of partial correlation matrices for all requested methods.
networks_corrected	Partial correlation matrix from naive EBICglasso (ignoring ceiling/floor effects).
networks_naive	Named list of naive partial correlation matrices (one per requested method), estimated without ceiling/floor correction.
Sigma_corrected	The corrected covariance matrix (p x p).
Sigma_naive	The raw sample covariance matrix (p x p).
diagnostics	List with per-variable censoring statistics, nearPD correction magnitude, and floor/ceiling thresholds. censoring Data frame (p rows) with per-variable censoring statistics: variable, n_floor, n_ceil, pct_floor, pct_ceil, naive and corrected means (mu_naive, mu_corrected), and standard deviations (sd_naive, sd_corrected). npd_correction Relative Frobenius-norm magnitude of the nearest positive-definite projection applied to the corrected covariance matrix. Values above 0.05 suggest instability and should be interpreted cautiously. a_vec Length-p numeric vector of effective floor thresholds used internally (with NULL inputs replaced by data-driven lower bounds). b_vec Length-p numeric vector of effective ceiling thresholds (with NULL inputs replaced by data-driven upper bounds).
method	Character vector of estimation method(s) run.
gamma	EBIC hyperparameter value used.
nlambda	Number of regularization parameters searched.
lambda.min.ratio	Smallest lambda as a fraction of the largest.
threshold	Logical; whether EBICglasso hard-thresholding was applied.
alpha	Significance level for p-value methods.
n	The number of observations.
p	The number of variables.
varnames	The character vector of variable (column) names.
data	Original data matrix.
call	Matched call.

Examples

# Simulate data with ceiling effects
set.seed(42)
Y <- MASS::mvrnorm(200, rep(0,5), diag(5) + 0.3)
Y_cens <- pmin(Y, 1.0)   # ceiling at 1.0 SD
fit <- ggm_cfe(Y_cens, floor=NULL, ceiling=1.0)
print(fit)
plot(fit)

# With psych::bfi personality data
# library(psych)
# fit <- ggm_cfe(bfi[,1:25], floor=1, ceiling=6)

---

induce.cfe

Description

inducing ceiling/floor effects in data

Usage

induce.cfe(floor.perc, ceiling.perc, y)

Arguments

floor.perc	a (non-empty) numeric value from 0 to 1 denoting the desired percentage of floor effects
ceiling.perc	a (non-empty) numeric value from 0 to 1 denoting the desired percentage of ceiling effects
y	a (non-empty) numeric vector of data

Value

y scores with induced ceiling/floor effects

Examples

x=rnorm(1000,0,1) #simulate "healthy data"
x.c20=induce.cfe(0,.2,x) #induce 20% ceiling effects into the data
sum(x.c20==max(x.c20))/length(x.c20) #check ceiling percentage
x.f20=induce.cfe(.2,0,x) #induce 20% floor effects into the data
sum(x.f20==min(x.f20))/length(x.f20) #check ceiling percentage

---

lw.f.star

Description

conduct an F star test for data with ceiling/floor effects

Usage

lw.f.star(data, formula, flr, ceil, method_type)

Arguments

data	a dataframe of data with ceiling/floor effects and corresponding group variables in wide format
formula	a formula denoting the dependent and independent variable, e.g., y~group
flr	a (non-empty) numeric value of the floor threshold (e.g., the minimum score of the measurement scale)
ceil	a (non-empty) numeric value of the ceiling threshold (e.g., the maximum score of the measurement scale)
method_type	a character string specifying the preferred method type. "a" uses the original sample size and "b" uses after-truncation sample size.

Value

statistic	the value of the Brown-Forsythe F star statistics
p.value	the p-value for the test
est.f.squared	effect size estimate in Cohen's f squared

Examples

dat=threeganova.sim(1000,.16,1)
dat[dat$group==1,3]=induce.cfe(0,.15,dat[dat$group==1,3])
lw.f.star(dat, y~group, flr=min(dat$y), ceil=max(dat$y), "a") #using truncated n
lw.f.star(dat, y~group, flr=min(dat$y), ceil=max(dat$y), "b") #using original n

---

lw.t.test

Description

conduct a t test adjusting for ceiling and/or floor effects

Usage

lw.t.test(x1, x2, flr1, ceil1, flr2, ceil2, method_type)

Arguments

x1	a (non-empty) numeric vector of data values for group 1 with floor/ceiling effects
x2	a (non-empty) numeric vector of data values for group 2 with floor/ceiling effects
flr1	a (non-empty) numeric value of the floor threshold for group 1
ceil1	a (non-empty) numeric value of the ceiling threshold for group 1
flr2	a (non-empty) numeric value of the floor threshold for group 2
ceil2	a (non-empty) numeric value of the ceiling threshold for group 2
method_type	a character string specifying the preferred method type. "a" uses the original sample size and "b" uses after-truncation sample size.

Value

statistic	the value of the adjusted t test statistics
p.value	the p-value for the test
est.d	effect size estimate as in Cohen's d
conf.int	95% confidence interval

Examples

x1.c=induce.cfe(0,.3,rnorm(1000,20,5)) #group 1 scores with 30% ceiling data
x2.c=induce.cfe(.15,0,rnorm(1000,30,5)) #group 2 scores with 15% floor data
lw.t.test(x1.c, x2.c, flr1=min(x1.c), ceil1=max(x1.c), flr2=min(x2.c), ceil2=max(x2.c), "a")
lw.t.test(x1.c, x2.c, flr1=min(x1.c), ceil1=max(x1.c), flr2=min(x2.c), ceil2=max(x2.c), "b")

---

Censoring Profile Plot for ggm_cfe Objects

Description

Plots the floor and ceiling proportions for each variable in a ggm_cfe object. Uses ggplot2 if available, otherwise falls back to base graphics.

Usage

plot_censoring(x)

Arguments

x	An object of class "ggm_cfe" as returned by ggm_cfe.

Value

Invisibly returns x.

Examples

set.seed(42)
Y <- MASS::mvrnorm(200, rep(0, 5), diag(5) + 0.3)
Y_cens <- pmin(Y, 1.0)
fit <- ggm_cfe(Y_cens, floor = NULL, ceiling = 1.0)
plot_censoring(fit)

---

Test dataset for DACF package

Description

A dataset used to demonstrate ceiling/floor effect corrections.

Usage

Qi_data_males

Format

A data frame with 1731 rows and 2 columns:

group

The values of the grouping variable. The participants were divided into three groups: moderate Problematic Internet Use (PIU) and Internet Gaming Disorder (IGD) risk, high PIU risk, and high IGD risk.

value

The observed data of the dependent variable, IGD severity from a scale from 0 to 9.

Source

Qi, H., Dong, Z., Wei, Q., Chen, X., & Luo, Y. (2025). Are Gamers Happier? Multidimensional Well-Being Differences in Risk Groups for Problematic Internet Use and Internet Gaming Disorder. Personality and Individual Differences 246, 113334. doi:10.2139/ssrn.5148047

---

rec.mean.var

Description

recover mean and variance of the data with ceiling/floor effects

Usage

rec.mean.var(y, flr, ceil)

Arguments

y	a (non-empty) numeric vector of data with ceiling/floor effects
flr	a (non-empty) numeric value of the floor threshold (e.g., the minimum score of the measurement scale)
ceil	a (non-empty) numeric value of the ceiling threshold (e.g., the maximum score of the measurement scale)

Value

ceiling.percentage	the percentage of ceiling values in the data
floor.percentage	the percentage of floor values in the data
est.mean	estimated mean of the true scores
est.var	estimated variance of the true scores

Examples

# simulate normally distributed true scores
x=rnorm(1000,2,4)
mean(x); var(x)
# induce 20% floor effects
# and estimate the true mean variance from the floor data
x.f=induce.cfe(.2,0,x)
rec.mean.var(x.f, flr=min(x.f), ceil=max(x.f))
# induce 20% ceiling effects
# and estimate the true mean and variance from the ceiling data
x.c=induce.cfe(0,.2,x)
rec.mean.var(x.c, flr=min(x.c), ceil=max(x.c))
# induce 20% and 10% of floor and ceiling effects, respectively
# and estimate the true mean and variance from the data with floor and ceiling effects
x.cf=induce.cfe(.2,.1,x)
rec.mean.var(x.cf, flr=min(x.cf), ceil=max(x.cf))

---

threeganova.sim

Description

simulate three-group anova data

Usage

threeganova.sim(group_n, f_sqr, sd.1)

Arguments

group_n	a (non-empty) numeric value of desired sample size per group
f_sqr	a (non-empty) numeric value of desired Cohen's f squared value
sd.1	a (non-empty) numeric value of desired standard deviation ratio

Value

a dataframe containing scores "y", grouping factor "group", and residual errors.

Examples

sample.3g=threeganova.sim(1000,.16,5) #data of n=1000, sd1=sd3=1 and sd2=5, and f^2=.16
colnames(sample.3g) #examine the column names
dim(sample.3g) #examine the data structure
aggregate(sample.3g$y,sd,by=list(sample.3g$group)) #check group standard deviations

---