Phenobarb Dataset

Description

This dataset contains longitudinal data on Phenobarbital concentration levels in newborn infants.

Usage

Phenobarb

Format

A data frame with 744 observations on the following 7 variables:

Subject

An ordered factor identifying the infant.

Wt

A numeric vector giving the birth weight of the infant (kg).

Apgar

An ordered factor giving the 5-minute Apgar score for the infant. This is an indication of the newborn's health.

ApgarInd

A factor indicating whether the 5-minute Apgar score is < 5 or >= 5.

time

A numeric vector giving the time when the sample is drawn or the drug is administered (hr).

dose

A numeric vector giving the dose of drug administered (\mu g/kg).

conc

A numeric vector giving the phenobarbital concentration in the serum (\mu g/L).

Source

MEMSS R package

Examples

data(Phenobarb)
head(Phenobarb)

Simulated Longitudinal Data

Description

This dataset consists of 10000 subjects with irregular observation times and influential observations.

Usage

infsdata

Format

A data frame contains:

subject_id

Unique identifier for each subject

time

Time points of observation

response

Simulated response value

subject_type

Category of subject (e.g., Influential, Non-Influential)

Source

Simulated dataset

Relative Longitudinal Difference (RLD)

Description

This function identifies influential subjects in longitudinal data based on their relative change in response over time. It helps in detecting subjects whose response values exhibit significant fluctuations beyond a specified threshold (k standard deviations).

Usage

rld(data, subject_id, time, response, k = 2, verbose = FALSE)

Arguments

Details

The function follows these steps:

• Computes the relative change in response values over time for each subject.

• Calculates the threshold based on k standard deviations from the mean relative change.

• Identifies subjects whose relative change exceeds the threshold.

• Separates data into influential and non-influential subjects.

• Generates visualizations to highlight influential subjects.

This method is particularly useful for detecting subjects with extreme response variations in longitudinal studies.

Value

A list containing:

See Also

tvm, wlm, sld, slm

Examples

data(infsdata)
infsdata <- infsdata[1:5,]
result <- rld(infsdata, "subject_id", "time", "response", k = 2)
print(result$influential_subjects)
head(result$influential_data)
head(result$non_influential_data)

Simple Longitudinal Difference (SLD)

Description

This function detects influential subjects in a longitudinal dataset by analyzing their successive differences. It calculates the successive differences for each subject, determines a threshold using the mean and standard deviation, and identifies subjects whose maximum successive difference exceeds this threshold. This approach helps in detecting abrupt changes in subject responses over time.

Usage

sld(data, subject_id, time, response, k = 2, verbose = FALSE)

Arguments

Details

The function follows these steps:

• Computes successive differences for each subject.

• Calculates the mean and standard deviation of these differences across all subjects.

• Defines a threshold as k standard deviations from the mean.

• Identifies subjects whose maximum successive difference exceeds this threshold.

• Separates data into influential and non-influential subjects.

• Visualizes the results using ggplot2.

This method is useful for identifying subjects with sudden changes in their response patterns over time.

Value

A list containing:

See Also

tvm, wlm, slm, rld

Examples

data(infsdata)
infsdata <- infsdata[1:5,]
result <- sld(infsdata, "subject_id", "time", "response", k = 2)
print(result$influential_subjects)
head(result$influential_data)
head(result$non_influential_data)

Simple Longitudinal Mean (SLM)

Description

This function detects influential subjects in longitudinal data based on their mean response values. It identifies subjects whose mean response deviates significantly beyond a specified threshold (defined as k standard deviations from the mean). The function provides a summary of influential subjects, separates the data into influential and non-influential subjects, calculates influence scores, and visualizes the results using ggplot2.

Usage

slm(data, subject_id, time, response, k = 2, verbose = FALSE)

Arguments

Details

The function follows these steps:

• Calculates the mean and standard deviation of the response variable across all subjects.

• Determines the threshold for influence based on k standard deviations from the mean.

• Identifies subjects whose mean response falls outside this threshold.

• Calculates the Influence Score (IS) for each subject as the absolute deviation of their mean from the overall mean.

• Calculates the Proportional Influence Score (PIS) for each subject as IS divided by the overall standard deviation.

• Separates data into influential and non-influential subjects.

• Visualizes the distribution of responses and highlights influential subjects.

This method is useful for detecting outliers and understanding the impact of extreme values in longitudinal studies.

Value

A list containing:

See Also

tvm, wlm, sld, rld

Examples

data(infsdata)
infsdata <- infsdata[1:5,]
result <- slm(infsdata, "subject_id", "time", "response", 2)
print(result$influential_subjects)
head(result$influential_data)
head(result$non_influential_data)
head(result$influence_scores)
print(result$mean_plot)
print(result$longitudinal_plot)

Time-Varying Mean (TVM)

Description

This function detects influential subjects based on their response values at different time points. It calculates the mean and standard deviation of responses at each time point and flags subjects whose response values deviate significantly beyond a threshold. The function also generates plots to visualize influential observations and their trends over time. It also computes the Influence Score (IS) and Partial Influence Score (PIS) for each observation.

Usage

tvm(data, subject_id, time, response, k = 2, verbose = FALSE)

Arguments

Details

The function follows these steps:

• Computes the mean and standard deviation of response values at each time point.

• Calculates Influence Score (IS) and Partial Influence Score (PIS) for each observation.

• Identifies subjects whose response values exceed the threshold based on k standard deviations.

• Separates influential and non-influential subjects for further analysis.

• Generates visualizations of mean responses and highlights influential subjects in a longitudinal plot.

This method is useful for identifying outliers and understanding variability in longitudinal studies.

Value

A list containing:

See Also

slm, wlm, sld, rld

Examples

data(infsdata)
infsdata <- infsdata[1:5,]
result <- tvm(infsdata, "subject_id", "time", "response", 2)
print(result$influential_subjects)
head(result$influential_data)
head(result$non_influential_data)
head(result$influential_time_data)
head(result$IS_table)
head(result$PIS_table)
result$mean_response_plot
result$longitudinal_plot

tvm.imputation: Impute Influential Responses in Longitudinal Data

Description

This function identifies influential response values using the 'tvm' function, replaces them with NA, and imputes the missing values using the 'mice' package.

Usage

tvm.imputation(
  data,
  subject_col,
  time_col,
  response_col,
  k,
  impute_method = "pmm",
  m = 5
)

Arguments

Value

A data frame with imputed values for the influential response points while maintaining original NA values.

Examples

infsdata <- infsdata[1:5,]
imptvm <- tvm.imputation(infsdata, "subject_id", "time", "response", k = 3)
head(imptvm)

Weighted Longitudinal Mean (WLM)

Description

This function identifies influential subjects in a longitudinal dataset based on their weighted mean response values. It computes weighted averages for each subject and detects anomalies by comparing them against an overall mean threshold.

Usage

wlm(data, subject_id, time, response, k = 2, verbose = FALSE)

Arguments

Details

The function follows these steps:

• Computes the weighted mean response for each subject.

• Calculates the overall mean and standard deviation of weighted responses.

• Identifies subjects whose weighted mean response deviates beyond k standard deviations.

• Separates data into influential and non-influential subjects.

• Provides visualizations of the detected anomalies using ggplot2.

This method is beneficial for detecting influential subjects in longitudinal studies, where responses may vary over time and require weighted adjustments.

Value

A list containing:

See Also

tvm, slm, sld, rld

Examples

data(infsdata)
infsdata <- infsdata[1:5,]
result <- wlm(infsdata, "subject_id", "time", "response", k = 2)
print(result$influential_subjects)
head(result$influential_data)
head(result$non_influential_data)
print(result$weighted_plot)
print(result$longitudinal_plot)