Type: Package
Title: Nonparametric Tests for Incomplete Clustered Data
Version: 0.1.1
Date: 2024-12-03
Maintainer: Yue Cui <YueCui@MissouriState.edu>
Description: Nonparametric tests for clustered data in pre-post intervention design documented in Cui and Harrar (2021) <doi:10.1002/bimj.201900310> and Harrar and Cui (2022) <doi:10.1016/j.jspi.2022.05.009>. Other than the main test results mentioned in the reference paper, this package also provides a function to calculate the sample size allocations for the input long format data set, and also a function for adjusted/unadjusted confidence intervals calculations. There are also functions to visualize the distribution of data across different intervention groups over time, and also the adjusted/unadjusted confidence intervals.
License: GPL-2 | GPL-3 [expanded from: GPL (≥ 2)]
Encoding: UTF-8
LazyData: true
Repository: CRAN
Date/Publication: 2024-12-03 07:20:02 UTC
RoxygenNote: 7.2.3
Depends: R (≥ 2.10)
Suggests: testthat (≥ 3.0.0)
Config/testthat/edition: 3
Imports: graphics, stats, MASS, ggplot2
NeedsCompilation: no
Packaged: 2024-12-03 07:05:04 UTC; yuecui
Author: Yue Cui ORCID iD [aut, cre], Solomon W. Harrar ORCID iD [aut]

Asthma Randomized Trial of Indoor-Wood Smoke Data

Description

A subset of the data set from a randomized trial of interventions to improve childhood asthma in homes with wood-burning stoves. The original data was collected on 115 children with asthma living in 98 eligible households. The outcomes are domain scores for Pediatric Asthma Quality of Life Questionnaire (PAQLQ) in symptoms, activity limitation and emotional function.

Usage

ARTIS

Format

A data frame with 50 rows and 6 variables:

homeid

unique id for each household

intervention

indicator for intervention, where 0 = pre-intervention, 1=after-intervention

tx

intervention type, where 1 = sham fiter, 2 = updated wood-burning stove, 3 = air-filter

symptoms_pqol

PAQLQ for symtoms

act_pqol

PAQLQ score for activity

emot_pqol

PAQLQ score for emotional function

Source

Noonan, Curtis W., and Tony J. Ward. "Asthma randomized trial of indoor wood smoke (ARTIS): rationale and methods." Contemporary clinical trials 33, no. 5 (2012): 1080-1087.

References

Noonan, Curtis W., Erin O. Semmens, Paul Smith, Solomon W. Harrar, Luke Montrose, Emily Weiler, Marcy McNamara, and Tony J. Ward. "Randomized trial of interventions to improve childhood asthma in homes with wood-burning stoves." Environmental health perspectives 125, no. 9 (2017): 097010. ([PubMed](https://pubmed.ncbi.nlm.nih.gov/28935614/));

Examples

data(ARTIS)
head(ARTIS)

Confidence Interval

Description

Construct confidence intervals for effect sizes.

Usage

ConfInterval(object, level, side="two.sided",
adjust=NULL)

Arguments

object

a fitted model object from ncda().

level

the confidence level required.

side

a character string specifying the side of the confidence bound, must be one of "two.sided" (default), "left" or "right".

adjust

an optional character string specifying the multiple adjustment method, by default there is no adjustment, if specified, must be one of "Bonferroni" or "Working-Hotelling". You can specify just the initial letter.

Value

A list or a vector. If the confidence interval is two-sided, lower and upper bounds are stored in lists for each nonparametric effect size estimate. Otherwise, the lower/upper bounds are stored in vectors in the order of the effect size estimates.

Examples

skin_analysis <- ncda(score~tx, skin, intervention, subject,
                      indicator=c("control","treatment"),
                      Contrast=matrix(c(1,-1), nrow = 1))
ConfInterval(skin_analysis,0.95)

ARTIS_analysis <- ncda(emot_pqol~tx, ARTIS, intervention, homeid,
                       indicator = c("0","1"))
ConfInterval(ARTIS_analysis,0.95)
ConfInterval(ARTIS_analysis,0.95,"two.sided","B")
ConfInterval(ARTIS_analysis,0.95,"left","W")

Bar plots for two-sided confidence intervals

Description

Bar plots for two-sided confidence intervals

Usage

Plot.ConfInt(object, level, side="two.sided",
adjust=NULL)

Arguments

object

a fitted model object from ncda().

level

the confidence level required.

side

a character string specifying the side of the confidence bound, must be one of "two.sided" (default), "left" or "right".

adjust

an optional character string specifying the multiple adjustment method, by default there is no adjustment, if specified, must be one of "Bonferroni" or "Working-Hotelling". You can specify just the initial letter.

Value

Bar plots.

Examples

skin_analysis <- ncda(score~tx, skin, intervention, subject,
                      indicator=c("control","treatment"),
                      Contrast=matrix(c(1,-1), nrow = 1))
Plot.ConfInt(skin_analysis,0.95,"Two-Sided")

ARTIS_analysis <- ncda(emot_pqol~tx, ARTIS, intervention, homeid,
                       indicator = c("0","1"))
Plot.ConfInt(ARTIS_analysis,0.95,"Two-Sided")
Plot.ConfInt(ARTIS_analysis,0.95,"Two-Sided","Bonferroni")
Plot.ConfInt(ARTIS_analysis,0.95,"Two-Sided","Working-Hotelling")

Box plots.

Description

Box plot of the input data set by treatments and time period.

Usage

Plot.box(object)

Arguments

object

a fitted model object from ncda() or a processed data set from ProcessData()

Value

Box plots.

Examples

#Plot from analysis object

ARTIS_analysis <- ncda(emot_pqol~tx, ARTIS, intervention, homeid,
                       indicator = c("0","1"))
Plot.box(ARTIS_analysis)

skin_analysis <- ncda(score~tx, skin, intervention, subject,
                      indicator = c("control","treatment"))
Plot.box(skin_analysis)
# Plot from processed data set
ARTIS_result <- ProcessData(ARTIS, tx, intervention, homeid, symptoms_pqol,
                            indicator = c("0","1"))
skin_result <- ProcessData(skin, tx, intervention, subject, score,
                           indicator = c("control","treatment"))
Plot.box(ARTIS_result)
Plot.box(skin_result)

Process data set.

Description

Sample size and cluster size calculation for the imported data set.

Usage

ProcessData(data, tx, period, subject, resp, indicator=NULL)

Arguments

data

a data frame in the long format.

tx

treatment variable.

period

time indicator variable.

subject

subject or cluster ID

resp

response variable to be analyzed.

indicator

an optional vector of characters indicating the order of pre and post intervention period; must match the levels of period argument if specified; if not specified, the pre and post intervention period will be ordered in the alphabet order by default

Value

a list containing the following components:

trt

number of treatments

nc

complete cluster sample size within each treatment group

n1

incomplete cluster sample size pre intervention within each treatment group

n2

incomplete cluster sample size post intervention within each treatment group

m1c

complete cluster sizes pre-intervention within each treatment group

m2c

complete cluster size post-intervention within each treatment group

m1i

incomplete cluster sizes pre-intervention within each treatment group

m2i

incomplete cluster sizes post-intervention within each treatment group

x1c

complete data pre-intervention within each treatment group

x2c

complete data post-intervention within each treatment group

x1i

incomplete data pre-intervention within each treatment group

x2i

incomplete data post-intervention within each treatment group

Examples

ARTIS_result <- ProcessData(ARTIS, tx, intervention, homeid, symptoms_pqol,
                            c("0","1"))
names(ARTIS_result)
skin_result <- ProcessData(skin, tx, intervention, subject, score,
                           c("control","treatment"))
skin_result$nc
skin_result$n1
skin_result$n2

Nonparametric Clustered Data Analysis

Description

Main function to calculate nonparametric effect sizes and conduct hypothesis tests.

Usage

ncda(formula,data,period,subject,indicator=NULL,Contrast=NULL)

Arguments

formula

An object of class "formula" (or one that can be coerced to that class): a symbolic description of the model to be fitted. The details of model specification are given under ‘Details’.

data

a data frame in the long format.

period

time indicator variable.

subject

subject or cluster ID

indicator

an optional vector of characters indicating the order of pre and post intervention period; must match the levels of period argument if specified; if not specified, the pre and post intervention period will be ordered in the alphabet order by default

Contrast

an optional contrast matrix for effect sizes.

Details

The model has the form response ~ tx where response is the (numeric) response variable and tx is the treatment variable.

Value

An object with effect sizes and other test details.

References

Cui, Yue, Frank Konietschke, and Solomon W. Harrar. "The nonparametric Behrens–Fisher problem in partially complete clustered data." Biometrical Journal 63.1 (2021): 148-167.

Harrar, Solomon W., and Yue Cui. "Nonparametric methods for clustered data in pre-post intervention design." Journal of Statistical Planning and Inference 222 (2023): 1-21.

Examples


ARTIS_analysis <- ncda(symptoms_pqol~tx, ARTIS, intervention, homeid,
                        indicator=c("0","1"),
                        Contrast=matrix(c(1,-1,1,-1,1,-1), nrow = 1))
names(ARTIS_analysis)
ARTIS_analysis$p.vector

skin_analysis <- ncda(score~tx, skin, intervention, subject,
                      indicator=c("control","treatment"),
                      Contrast=matrix(c(1,-1), nrow = 1))
skin_analysis$TotalSampleSize
skin_analysis$p.vector

Skin Irritation Data

Description

The data set is a re-simulated part from an ongoing neuodermatitis study where researchers investigate the efficacy of an ointment in reducing the severity of skin irritation on the backs of the hands of 25 neurodermatitis patients, where 10 patients’ backs of the hands were rubbed with the ointment and 15 were not. The response is a BI-RADS rating score and the lower the score the better the clinical record. Every remarkable skin irritation was graded on every patients back of the hands and thus, the numbers of replicates differ across the patients.

Usage

skin

Format

A data frame with 107 rows and 4 variables:

tx

treatment group

intervention

intervention period indicator

subject

subject ID

score

BI-RADS rating score, where 1 = very mild irritation, 2 =slight irritation, 3 =mild irritation, 4 =heavy irritation and 5 =severe irritation

References

Roy, A, Harrar, SW, Konietschke, F. The nonparametric Behrens-Fisher problem with dependent replicates. Statistics in Medicine. 2019; 38: 4939– 4962. https://doi.org/10.1002/sim.8343

Examples

data(skin)
head(skin)