| Title: | Tools for Simulating Direct Behavioral Observation Recording Procedures Based on Alternating Renewal Processes |
|---|---|
| Description: | Tools for simulating data generated by direct observation recording. Behavior streams are simulated based on an alternating renewal process, given specified distributions of event durations and interim times. Different procedures for recording data can then be applied to the simulated behavior streams. Functions are provided for the following recording methods: continuous duration recording, event counting, momentary time sampling, partial interval recording, whole interval recording, and augmented interval recording. |
| Authors: | James E. Pustejovsky, with contributions from Daniel M. Swan |
| Maintainer: | James E. Pustejovsky <[email protected]> |
| License: | GPL-3 |
| Version: | 1.2.2 |
| Built: | 2025-02-15 03:24:03 UTC |
| Source: | https://github.com/jepusto/arpobservation |
Tools for simulating different methods of observing alternating renewal processes
ARPobservation provides a set of tools for simulating data based on direct observation of behavior. It works by first simulating a behavior stream based on an alternating renewal process, using specified distributions of event durations and interim times. Different procedures for recording data can then be applied to the simulated behavior stream.
The main function for simulating a behavior stream is r_behavior_stream. Currently, the event
duration and interim time distributions must come from the class eq_dist. (See the documentation
for this class for distributions that are currently implemented.)
Several different observation recording procedures can then be applied as filters to a simulated behavior stream. The following procedures are currently implemented:
To apply multiple procedures to the same behavior stream, use reported_observations. Data can also
be simulated using the convenience functions r_PIR, r_WIR, r_MTS,
r_continuous_recording, and r_event_counting. These functions wrap the
behavior-stream generation step and the observation recording step into a single function. They are more
memory efficient, but slightly less computationally efficient, than executing each step in turn.
James E. Pustejovsky <[email protected]>
An interactive tool that simulates single-case designs with outcomes measured by systematic direct observation. The behavioral outcomes are generated from the alternating renewal process model. Both event behaviors and state behaviors are supported. Event behaviors are generated from a renewal process with gamma-distributed inter-event times. State behaviors are generated from an alternating poisson process, in which the event durations and interim times are exponentially distributed. Currently, multiple baseline and treatment reversal designs are supported.
ARPsimulator(launch_browser = TRUE)ARPsimulator(launch_browser = TRUE)
launch_browser |
Logical value indicating whether to run the tool using
the system's default web browser. Defaults to |
Divides the observation session into intervals. Each interval is scored using partial interval recording, whole interval recording, and momentary time sampling (at the beginning of the following interval).
augmented_recording(BS, interval_length, rest_length = 0)augmented_recording(BS, interval_length, rest_length = 0)
BS |
object of class behavior_stream |
interval_length |
time length of each interval. |
rest_length |
portion of each interval to exclude from observation. Default is zero. See details. |
Each behavior stream is divided into intervals of length interval_length.
The last rest_length of each interval is excluded from observation.
For example, for a stream length of 100, interval_length = 20, and
rest_length = 5, the first interval runs from [0,15), the second interval runs from [20,35), etc.
A matrix with rows equal to the number of intervals per session and
columns equal to the number of behavior streams in BS.
BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) augmented_recording(BS, interval_length = 20)BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) augmented_recording(BS, interval_length = 20)
Calculates the proportion of session time during which behavior occurs.
continuous_duration_recording(BS)continuous_duration_recording(BS)
BS |
object of class |
Vector of proportions.
BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) continuous_duration_recording(BS)BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) continuous_duration_recording(BS)
Single case design data measured with partial interval recording from a study of the effect of providing Choice between academic activities on the disruptive behavior of three elementary school students with emotional and behavioral disorders. For this data "No Choice" is the baseline phase. Data were extracted from the figures in the publication.
A data frame with 58 observations on 7 variables
Case The participant for whom the observation took place
Phase The level of the observation ("Choice" vs. "No Choice")
Session The observation session # for each participant
outcome The summary PIR measurement for the observation session
active_length The length of the active observation interval, in seconds
rest_length The length of the recording interval, in seconds
intervals The total number of intervals in the observation session
A data frame with 53940 rows and 10 variables
Dunlap, G., DePerczel, M., Clarke, S., Wilson, D., Wright, S., White, R., & Gomez, A. (1994). Choice making to promote adaptive behavior for students with emotional and behavioral challenges. Journal of Applied Behavior Analysis, 27 (3), 505-518.
eq_dist
The eq_dist class consists of a pair of component functions for generating random variates from a
specified distribution and the corresponding equilibrium distribution.
eq_dist(r_gen, r_eq)eq_dist(r_gen, r_eq)
r_gen |
function for generating random deviates. |
r_eq |
function for generating random deviates from the corresponding equilibrium distribution. |
Both functions must take arguments n and mean. Currently, the following distributions
are implemented:
F_exp - Exponential
F_gam - Gamma
F_gam_mix - Mixture of two gammas
F_weib - Weibull
F_unif - Uniform
F_const - Constant
Object of class eq_dist with components r_gen and r_eq.
Calculates the number of behaviors that begin during the observation session.
event_counting(BS)event_counting(BS)
BS |
object of class |
Vector of non-negative integers.
BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) event_counting(BS)BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) event_counting(BS)
Generation from a degenerate distribution and random number generation from the related equilibrium distribution,
for use with r_behavior_stream.
F_const()F_const()
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) simply returns a vector of length n with all values equal to mean.
The function r_eq(n, mean) generates random deviates from a uniform distribution on the interval (0, mean).
hist(F_const()$r_gen(1000, 2)) hist(F_const()$r_eq(1000, 2))hist(F_const()$r_gen(1000, 2)) hist(F_const()$r_eq(1000, 2))
Random number generation from exponential distributions, for use with r_behavior_stream.
F_exp()F_exp()
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) generates random deviates from an exponential distribution with specified mean.
The function r_eq(n, mean) generates random deviates from an exponential distribution with specified mean.
hist(F_exp()$r_gen(1000, 3)) hist(F_exp()$r_eq(1000, 3))hist(F_exp()$r_gen(1000, 3)) hist(F_exp()$r_eq(1000, 3))
Random number generation from a gamma distribution and the related equilibrium distribution,
for use with r_behavior_stream.
F_gam(shape)F_gam(shape)
shape |
shape parameter |
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) generates random deviates from a gamma distribution with specified
mean and shape parameters.
The function r_eq(n, mean) generates random deviates from the equilibrium distribution corresponding
to the gamma distribution with specified mean and shape parameters.
hist(F_gam(2)$r_gen(1000, 3)) hist(F_gam(2)$r_eq(1000, 3))hist(F_gam(2)$r_gen(1000, 3)) hist(F_gam(2)$r_eq(1000, 3))
Random number generation from a mixture of two gamma distributions and the related equilibrium distribution,
for use with r_behavior_stream.
F_gam_mix(shape1, shape2, scale_ratio, mix)F_gam_mix(shape1, shape2, scale_ratio, mix)
shape1 |
shape parameter for first mixture component, |
shape2 |
shape parameter for second mixture component, |
scale_ratio |
ratio of first scale component to second scale component, |
mix |
mixing proportion of first component, |
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) generates random deviates from a mixture of two gamma distributions with specified
mean, shape1, shape2, scale_ratio, and mix. The cumulative distribution function
is given by
where
is the cumulative distribution function of a Gamma random variable with shape and scale , and
the scale parameters are determined by the specified mean and scale_ratio.
The function r_eq(n, mean) generates random deviates from the equilibrium distribution corresponding
to the mixture of gamma distributions.
hist(F_gam_mix(2, 2, 1 / 12, 3 / 5)$r_gen(1000, 20)) hist(F_gam_mix(2, 2, 1 / 12, 3 / 5)$r_eq(1000, 20))hist(F_gam_mix(2, 2, 1 / 12, 3 / 5)$r_gen(1000, 20)) hist(F_gam_mix(2, 2, 1 / 12, 3 / 5)$r_eq(1000, 20))
Random number generation from a uniform distribution and the related equilibrium distribution,
for use with r_behavior_stream.
F_unif()F_unif()
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) generates random deviates from a uniform distribution with specified
mean on the interval . The cumulative distribution function
is given by .
The function r_eq(n, mean) generates random deviates from the equilibrium distribution corresponding
to a uniform distribution on the interval . The cumulative distribution function is given by
hist(F_unif()$r_gen(1000, 2)) hist(F_unif()$r_eq(1000, 2))hist(F_unif()$r_gen(1000, 2)) hist(F_unif()$r_eq(1000, 2))
Random number generation from a Weibull distribution and the related equilibrium distribution,
for use with r_behavior_stream.
F_weib(shape)F_weib(shape)
shape |
shape parameter |
Object of class eq_dist with components r_gen and r_eq.
The function r_gen(n, mean) generates random deviates from a Weibull distribution with specified
mean and shape parameters.
The function r_eq(n, mean) generates random deviates from the equilibrium distribution corresponding
to the Weibull distribution with specified mean and shape parameters.
hist(F_gam(2)$r_gen(1000, 3)) hist(F_gam(2)$r_eq(1000, 3))hist(F_gam(2)$r_gen(1000, 3)) hist(F_gam(2)$r_eq(1000, 3))
Calculates a bound for the log of the incidence ratio of two samples (referred to as baseline and treatment) based on partial interval recording (PIR) data, assuming that the behavior follows an Alternating Renewal Process.
incidence_bounds( PIR, phase, base_level, mu_U, p, active_length, intervals = NA, conf_level = 0.95, exponentiate = FALSE )incidence_bounds( PIR, phase, base_level, mu_U, p, active_length, intervals = NA, conf_level = 0.95, exponentiate = FALSE )
PIR |
vector of PIR measurements |
phase |
factor or vector indicating levels of the PIR measurements. |
base_level |
a character string or value indicating the name of the baseline level. |
mu_U |
the upper limit on the mean event duration |
p |
upper limit on the probability that the interim time between behavioral events is less than the active interval |
active_length |
length of the active observation interval |
intervals |
the number of intervals in the sample of observations. Default is |
conf_level |
Coverage rate of the confidence interval. Default is |
exponentiate |
Logical value indicating if the log of the bounds and the confidence interval should be exponentiated. Default is |
The incidence ratio estimate is based on the assumptions that
1) the underlying behavior stream follows an Alternating Renewal Process,
2) the average event duration is less than mu_U, and
3) the probability of observing an interim time less than the active interval length is less than p.
The PIR vector can be in any order corresponding to the factor or vector phase. The levels of phase can be any two levels, such as "A" and "B", "base" and "treat", or "0" and "1". If there are more than two levels in phase this function will not work. A value for base_level must be specified - if it is a character string it is case sensitive.
For all of the following variables, the function assumes that if a vector of values is provided they are constant across all observations and simply uses the first value in that vector.
mu_U is the upper limit on the mean event durations. This is a single value assumed to hold for both samples of behavior
active_length This is the total active observation length. If the intervals are 15 seconds long but 5 seconds of each interval is reserved for recording purposes, active_length= 10. Times are often in seconds, but can be in any time unit.
intervals is the number of intervals in the observations. This is a single value and is assumed to be constant across both samples and all observations. This value is only relevant if the mean of one of the samples is at the floor or ceiling of 0 or 1. In that case it will be used to truncate the sample mean. If the sample mean is at the floor or ceiling and no value for intervals is provided, the function will stop.
A list containing two named vectors and a single named number.
The first entry, estimate_bounds, contains the lower and upper bound for the estimate of the incidence ratio.
The second entry, estimate_SE, contains the standard error of the estimate.
The third entry, estimate_CI, contains the lower and upper bounds for the confidence interval of the incidence ratio.
Daniel Swan <[email protected]>
# Estimate bounds on the incidence ratio for Ahmad from the Dunlap dataset data(Dunlap) with(subset(Dunlap, Case == "Ahmad"), incidence_bounds(PIR = outcome, phase = Phase, base_level = "No Choice", mu_U = 10, p = .15, active_length = active_length, intervals = intervals))# Estimate bounds on the incidence ratio for Ahmad from the Dunlap dataset data(Dunlap) with(subset(Dunlap, Case == "Ahmad"), incidence_bounds(PIR = outcome, phase = Phase, base_level = "No Choice", mu_U = 10, p = .15, active_length = active_length, intervals = intervals))
Calculates a bound for the log of the ratio of interim time of two samples (referred to as baseline and treatment) based on partial interval recording (PIR) data, assuming that the average event durations are equal across samples and that interim times are exponentially distributed.
interim_bounds( PIR, phase, base_level, conf_level = 0.95, intervals = NA, exponentiate = FALSE )interim_bounds( PIR, phase, base_level, conf_level = 0.95, intervals = NA, exponentiate = FALSE )
PIR |
vector of PIR measurements |
phase |
factor or vector indicating levels of the PIR measurements. |
base_level |
a character string or value indicating the name of the baseline level. |
conf_level |
Desired coverage rate of the calculated confidence interval. Default is |
intervals |
the number of intervals in the sample of observations. Default is |
exponentiate |
Logical value indicating if the log of the bounds and the confidence interval should be exponentiated. Default is |
The interim ratio estimate is based on the assumptions that 1) the underlying behavior stream follows an Alternating Renewal Process, 2) the average event durations in each sample are equal, and 3) interim times follow exponential distributions.
The PIR vector can be in any order corresponding to the factor or vector phase.
The levels of phase can be any two levels, such as "A" and "B", "base" and "treat",
or "0" and "1". If there are more than two levels in phase this function will not work.
A value for base_level must be specified; if it is a character string it is case sensitive.
intervals is the number of intervals in the observations.
This is a single value and is assumed to be constant across both samples and all observations.
If intervals is sent as a vector instead of a single value, the first value in the vector will
be used. This value is only relevant if the mean of one of the samples is at the floor or ceiling
of 0 or 1. In that case it will be used to truncate the sample mean. If the sample mean is at the
floor or ceiling and no value for intervals is provided, the function will stop.
A list with three named entries
The first entry, estimate_bounds, contains the lower and upper bound for the estimate of the prevalence ratio.
The second entry, estimate_SE, contains the standard errors for the upper and lower bounds.
The third entry, estimate_CI, contains the lower and upper bounds for the confidence interval of the prevalence ratio.
Daniel Swan <[email protected]>
# Estimate bounds on the interim time ratio for Carl from the Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), interim_bounds(PIR = outcome, phase = Phase, base_level = "No Choice"))# Estimate bounds on the interim time ratio for Carl from the Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), interim_bounds(PIR = outcome, phase = Phase, base_level = "No Choice"))
Divides the observation session into a specified number of intervals. For partial interval recording, each interval is scored according to whether the behavior is present at any point during the interval. For whole interval recording, each interval is scored according to whether the behavior is present for the duration.
interval_recording( BS, interval_length, rest_length = 0, partial = TRUE, summarize = TRUE )interval_recording( BS, interval_length, rest_length = 0, partial = TRUE, summarize = TRUE )
BS |
object of class behavior_stream |
interval_length |
time length of each interval. |
rest_length |
portion of each interval to exclude from observation. Default is zero. See details. |
partial |
logical value indicating whether to use partial interval recording ( |
summarize |
logical value indicating whether vector of moments should be summarized by taking their mean. |
Each behavior stream is divided into intervals of length interval_length.
The last rest_length of each interval is excluded from observation.
For example, for a stream length of 100, interval_length = 20, and
rest_length = 5, the first interval runs from [0,15), the second interval runs from [20,35), etc.
If summarize = FALSE, a matrix with rows equal to the number of intervals per session and
columns equal to the number of behavior streams in BS.
If summarize = TRUE, a vector of proportions of length equal to the
number of behavior streams in BS.
BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) interval_recording(BS, interval_length = 20, partial = TRUE, summarize = FALSE) interval_recording(BS, interval_length = 20, partial = TRUE, summarize = TRUE) colMeans(interval_recording(BS, 20, partial = TRUE, summarize = FALSE)) interval_recording(BS, interval_length = 20, rest_length = 5, partial = FALSE)BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) interval_recording(BS, interval_length = 20, partial = TRUE, summarize = FALSE) interval_recording(BS, interval_length = 20, partial = TRUE, summarize = TRUE) colMeans(interval_recording(BS, 20, partial = TRUE, summarize = FALSE)) interval_recording(BS, interval_length = 20, rest_length = 5, partial = FALSE)
Estimates the log-response ratio (with or without bias correction), the variance of the log-response ratio, and the confidence interval for a given confidence level.
logRespRatio( observations, phase, base_level, conf_level = 0.95, bias_correct = TRUE, exponentiate = FALSE )logRespRatio( observations, phase, base_level, conf_level = 0.95, bias_correct = TRUE, exponentiate = FALSE )
observations |
Vector of observations |
phase |
Factor or vector indicating levels of the PIR measurements. |
base_level |
a character string or value indicating the name of the baseline level. |
conf_level |
Desired coverage rate of the calculated confidence interval. Default is |
bias_correct |
Logical value indicating if the bias-corrected log-response ratio should be used. Default is |
exponentiate |
Logical value indicating if the log-response ratio should be exponentiated. |
The observations vector can be in any order corresponding to the factor or vector phase.
The levels of phase can be any two levels, such as "A" and "B", "base" and "treat", or "0" and "1".
If there are more than two levels in phase this function will not work.
A value for base_level must be specified - if it is a character string it is case sensitive.
If exponentiate = TRUE, the log-ratio and the confidence interval will be exponentiated,
but the variance will be excluded from the output.
If exponentiate = FALSE, a list with three named entries.
The first entry, lRR, is the estimated log-response ratio.
The second entry, V_lRR, is the estimated variance of the log-response ratio.
The third entry, CI, is a vector containing the endpoints of a confidence
interval of conf_level coverage rate.
If exponentiate = TRUE, a list with two named entries.
The first entry, RR, is the estimated response ratio.
The second entry, CI, is a vector containing the endpoints of a confidence
interval of conf_level coverage rate.
Daniel Swan <[email protected]>
# Estimate the log response ratio and its variance for Carl from Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), logRespRatio(observations = outcome, phase = Phase, base_level = "No Choice"))# Estimate the log response ratio and its variance for Carl from Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), logRespRatio(observations = outcome, phase = Phase, base_level = "No Choice"))
Single-case design data from a study that using partial interval recording (PIR) examining the impact of choice-making in a homework tutoring context on disruptive behavior. In this data "No Choice" is the baseline phase. Data were extracted from the figure in the publication.
A data frame with 80 observations on 7 variables
Case The participant for whom the observation took place
Phase The level of the observation ("Choice" vs. "No Choice")
Session The observation session # for each participant
outcome The summary PIR measurement for the observation session
active_length The length of the active observation interval, in seconds
rest_length The length of the recording interval, in seconds
intervals The total number of intervals in the observation session
A data frame with 53940 rows and 10 variables
Moes, D. R. (1998). Integrating choice-making opportunities within teacher-assigned academic tasks to facilitate the performance of children with autism. Research and Practice for Persons with Severe Disabilities, 23 (4), 319-328.
Evaluates the presence or absence of the behavior at fixed moments in time.
momentary_time_recording(BS, interval_length, summarize = TRUE)momentary_time_recording(BS, interval_length, summarize = TRUE)
BS |
object of class behavior_stream |
interval_length |
length of interval between moments. |
summarize |
logical value indicating whether vector of moments should be summarized by taking their mean. |
If summarize = FALSE, a matrix with length n_intervals + 1 and width equal to the number
of behavior streams in BS. If summarize = TRUE, a vector of proportions of length equal to the
number of behavior streams in BS. Note that if summarize = TRUE, the initial state of the
behavior stream is excluded when calculating the mean, so the proportion is based on n_intervals values.
BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) momentary_time_recording(BS, interval_length = 20, FALSE) momentary_time_recording(BS, interval_length = 20) colMeans(momentary_time_recording(BS, 20, FALSE)[-1,])BS <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) momentary_time_recording(BS, interval_length = 20, FALSE) momentary_time_recording(BS, interval_length = 20) colMeans(momentary_time_recording(BS, 20, FALSE)[-1,])
Calculates the log-likelihood of within-session PIR data
PIR_loglik(param, U, c, d)PIR_loglik(param, U, c, d)
param |
vector of parameter values logit(prevalence), log(incidence) |
U |
a vector containing interval-level PIR data |
c |
the length of the active interval |
d |
the length of the recording interval The vector |
The value of the log-likelihood
Daniel Swan <[email protected]>
Estimates prevalence and incidence for two samples, along with the ratios of each parameter, assuming that the behavior follows an '. Also provides bootstrap confidence intervals.
PIR_MOM( PIR, phase, base_level, intervals, interval_length, rest_length = 0, Bootstraps = 2000, conf_level = 0.95, exponentiate = FALSE, seed = NULL )PIR_MOM( PIR, phase, base_level, intervals, interval_length, rest_length = 0, Bootstraps = 2000, conf_level = 0.95, exponentiate = FALSE, seed = NULL )
PIR |
vector of PIR measurements |
phase |
factor or vector indicating levels of the PIR measurements. |
base_level |
a character string or value indicating the name of the baseline level. |
intervals |
the number of intervals in the sample of observations |
interval_length |
the total length of each interval |
rest_length |
length of the portion of the interval devoted to recording. Default is |
Bootstraps |
desired number of bootstrap replicates. Default is |
conf_level |
Desired coverage rate of the calculated confidence interval. Default is |
exponentiate |
a logical indicating whether the row corresponding to the ratio of treatment to baseline should be exponentiated, with the default as |
seed |
seed value set in order to make bootstrap results reproducible. Default is |
The moment estimators are based on the assumption that the underlying behavior stream follows an Alternating Poisson Process, in which both the event durations and interim times are exponentially distributed.
The PIR vector can be in any order corresponding to the factor or vector
phase. The levels of phase can be any two levels, such as "A" and "B",
"base" and "treat", or "0" and "1". If there are more than two levels in phase
this function will not work. A value for base_level must be specified - if it is a
character string it is case sensitive.
intervals, interval_length, and rest_length are all single values that
are assumed to be held constant across both samples and all observation sessions.
If vectors of values are provided for these variables, it is assumed that the first value
in each vector is constant across all observations.
interval_length This is the total length of each individual interval.
Sometimes a portion of the interval is set aside for recording purposes, in which case
rest_length should be set to the length of time devoted to recording.
The default assumption is that there is no recording time.
The length of time devoted to active observation is calculated as interval_length - rest_length.
At the default setting of bootstraps = 2000, PIR_MOM takes just under six seconds to run on an Intel Core i5-2410M processor.
A dataframe with six columns and three rows corresponding to baseline, treatment,
and the log ratio or ratio (depending upon the value of exponentiate) of treatment to baseline
Daniel Swan <[email protected]>
# Estimate prevalence and incidence ratios for Carl from the Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), PIR_MOM(PIR = outcome, phase = Phase, intervals = intervals, interval_length = (active_length + rest_length), rest_length = rest_length, base_level = "No Choice", Bootstraps = 200, seed = 149568373))# Estimate prevalence and incidence ratios for Carl from the Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), PIR_MOM(PIR = outcome, phase = Phase, intervals = intervals, interval_length = (active_length + rest_length), rest_length = rest_length, base_level = "No Choice", Bootstraps = 200, seed = 149568373))
behavior_stream objectsCreates a graphical representation of a set of simulated behavior streams.
## S3 method for class 'behavior_stream' plot( x, session_color = "black", episode_color = "blue", episode_thickness = 2, ... )## S3 method for class 'behavior_stream' plot( x, session_color = "black", episode_color = "blue", episode_thickness = 2, ... )
x |
object of class |
session_color |
character string indicating the color of the lines that represent session time. Default is black. |
episode_color |
character string indicating the color of the bars that represent episode durations. Default is blue. |
episode_thickness |
numeric value indicating the thickness of the bars that represent episode durations. Default is 2. |
... |
Further arguments, not used for this method. |
The plot is created using ggplot from the ggplot2 package,
which must be installed.
An object of class ggplot.
if (requireNamespace("ggplot2", quietly = TRUE)) { b_streams <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) plot(b_streams) }if (requireNamespace("ggplot2", quietly = TRUE)) { b_streams <- r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) plot(b_streams) }
Calculates a bound for the log of the prevalence ratio of two samples (referred to as baseline and treatment) based on partial interval recording (PIR) data, assuming that the behavior follows an Alternating Renewal Process.
prevalence_bounds( PIR, phase, base_level, mu_L, active_length, intervals = NA, conf_level = 0.95, exponentiate = FALSE )prevalence_bounds( PIR, phase, base_level, mu_L, active_length, intervals = NA, conf_level = 0.95, exponentiate = FALSE )
PIR |
vector of PIR measurements |
phase |
factor or vector indicating levels of the PIR measurements. |
base_level |
a character string or value indicating the name of the baseline level. |
mu_L |
the lower limit on the mean event duration |
active_length |
length of the active observation interval |
intervals |
the number of intervals in the sample of observations. Default is |
conf_level |
Coverage rate of the confidence interval. Default is |
exponentiate |
Logical value indicating if the log of the bounds and the confidence interval should be exponentiated. Default is |
The prevalence ratio estimate is based on the assumptions that
1) the underlying behavior stream follows an Alternating Renewal Process and
2) the average event duration is greater than mu_L.
The PIR vector can be in any order corresponding to the factor or vector phase.
The levels of phase can be any two levels, such as "A" and "B", "base" and "treat", or "0" and "1".
If there are more than two levels in phase this function will not work.
A value for base_level must be specified - if it is a character string it is case sensitive.
For all of the following variables, the function assumes that if a vector of values is provided they are constant across all observations and simply uses the first value in that vector.
mu_L is the lower limit on the mean event durations. This is a single value assumed to hold for both samples of behavior
active_length This is the total active observation length. If the intervals are 15 seconds long but 5 seconds of each interval is reserved for recording purposes, active_length= 10. Times are often in seconds, but can be in any time unit.
intervals is the number of intervals in the observations. This is a single value and is assumed to be constant across both samples and all observations. This value is only relevant if the mean of one of the samples is at the floor or ceiling of 0 or 1. In that case it will be used to truncate the sample mean. If the sample mean is at the floor or ceiling and no value for intervals is provided, the function will stop.
A list with three named entries.
The first entry, estimate_bounds, contains the lower and upper bound for the estimate of the prevalence ratio.
The second entry, estimate_SE, contains the standard error of the estimate.
The third entry, estimate_CI, contains the lower and upper bounds for the confidence interval of the prevalence ratio.
Daniel Swan <[email protected]>
# Estimate bounds on the prevalence ratio for Carl from Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), prevalence_bounds(PIR = outcome, phase = Phase, base_level = "No Choice", mu_L = 10, active_length = active_length, intervals = intervals))# Estimate bounds on the prevalence ratio for Carl from Moes dataset data(Moes) with(subset(Moes, Case == "Carl"), prevalence_bounds(PIR = outcome, phase = Phase, base_level = "No Choice", mu_L = 10, active_length = active_length, intervals = intervals))
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, which are then coded as augmented interval recording data with given interval length and rest length.
r_AIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, equilibrium = TRUE, p0 = 0, tuning = 2 )r_AIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
interval_length |
total interval length |
rest_length |
length of any recording time in each interval |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies an augmented interval recording filter to the generated behavior streams.
An array with rows corresponding the number of intervals per session,
columns corresponding to MTS, PIR, and WIR records, all replicated n times.
n times.
James Pustejovsky <[email protected]>
r_AIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0)r_AIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0)
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution.
r_behavior_stream( n, mu, lambda, F_event, F_interim, stream_length, equilibrium = TRUE, p0 = 0, tuning = 2 )r_behavior_stream( n, mu, lambda, F_event, F_interim, stream_length, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
vector of mean event durations |
lambda |
vector of mean interim time |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
stream_length |
length of behavior stream |
equilibrium |
logical; if |
p0 |
vector of initial state probabilities. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and
random interim times from the distributions as specified, until the sum of the durations and interim
times exceeds the requested stream length. The vectors mu, lambda, and p0 are
recycled to length n.
An object of class behavior_stream containing two elements.
# default equilibrium initial conditions r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) # non-equilibrium initial conditions r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_gam(3), F_interim = F_gam(3), stream_length = 100, equilibrium = FALSE, p0 = 0.5)# default equilibrium initial conditions r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) # non-equilibrium initial conditions r_behavior_stream(n = 5, mu = 3, lambda = 10, F_event = F_gam(3), F_interim = F_gam(3), stream_length = 100, equilibrium = FALSE, p0 = 0.5)
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, summarized as the total proportion of time the behavior of interest occurred.
r_continuous_recording( n, mu, lambda, stream_length, F_event, F_interim, equilibrium = TRUE, p0 = 0, tuning = 2 )r_continuous_recording( n, mu, lambda, stream_length, F_event, F_interim, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies a continuous recording filter to the generated behavior streams.
A vector of proportions of length n.
Daniel Swan <[email protected]>
r_continuous_recording(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp())r_continuous_recording(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp())
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, summarized as the the total number of behaviors that began during the recording session
r_event_counting( n, mu, lambda, stream_length, F_event, F_interim, equilibrium = TRUE, p0 = 0, tuning = 2 )r_event_counting( n, mu, lambda, stream_length, F_event, F_interim, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies an event counting filter to the generated behavior streams.
A vector of behavior counts of length n.
Daniel Swan <[email protected]>
r_event_counting(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp())r_event_counting(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp())
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, which are then coded as momentary time sampling data with given interval length between moments.
r_MTS( n, mu, lambda, stream_length, F_event, F_interim, interval_length, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )r_MTS( n, mu, lambda, stream_length, F_event, F_interim, interval_length, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
interval_length |
length of time between moments |
summarize |
logical value indicating whether the vector of moments should be summarized by taking their mean, excluding the first moment in each row. |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time. |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies a momentary time sampling filter to the generated behavior streams.
If summarize = FALSE, a matrix of logicals with rows equal to n and length equal to (stream_length/interval_length) + 1. If summarize = TRUE, a vector of means of length n.
Daniel Swan <[email protected]>
# A set of unsummarized MTS observations r_MTS(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1) # A set of summarized MTS observations r_MTS(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, summarize = TRUE)# A set of unsummarized MTS observations r_MTS(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1) # A set of summarized MTS observations r_MTS(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, summarize = TRUE)
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, which are then coded as partial interval recording data with given interval length and rest length.
r_PIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )r_PIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
interval_length |
total interval length |
rest_length |
length of any recording time in each interval |
summarize |
logical value indicating whether the behavior streams should by summarized by taking their mean |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies a partial interval recording filter to the generated behavior streams.
If summarize = FALSE, a matrix with rows equal to n and a number of columns equal to the number intervals per session. If summarize = TRUE a vector of means of length n.
Daniel Swan <[email protected]>
# An unsummarized set of PIR observations r_PIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0) # A summarized set of of PIR observations r_PIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0, summarize = TRUE)# An unsummarized set of PIR observations r_PIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0) # A summarized set of of PIR observations r_PIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0, summarize = TRUE)
Random generation of behavior streams (based on an alternating renewal process) of a specified length and with specified mean event durations, mean interim times, event distribution, and interim distribution, which are then coded as whole interval recording data with given interval length and rest length.
r_WIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )r_WIR( n, mu, lambda, stream_length, F_event, F_interim, interval_length, rest_length = 0, summarize = FALSE, equilibrium = TRUE, p0 = 0, tuning = 2 )
n |
number of behavior streams to generate |
mu |
mean event duration |
lambda |
mean interim time |
stream_length |
length of behavior stream |
F_event |
distribution of event durations. Must be of class |
F_interim |
distribution of interim times. Must be of class |
interval_length |
total interval length |
rest_length |
length of any recording time in each interval |
summarize |
logical value indicating whether the behavior streams should by summarized by taking their mean |
equilibrium |
logical; if |
p0 |
Initial state probability. Only used if |
tuning |
controls the size of the chunk of random event durations and interim times. Adjusting this may be useful in order to speed computation time . |
Generates behavior streams by repeatedly drawing random event durations and random interim times from the distributions as specified, until the sum of the durations and interim times exceeds the requested stream length. Then applies a whole interval recording filter to the generated behavior streams.
If summarize = FALSE, a matrix with rows equal to n and a number of columns equal to the number intervals per session. If summarize = TRUE a vector of means of length n.
Daniel Swan <[email protected]>
# An unsummarized set of WIR observations r_WIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0) # A summarized set of of WIR observations r_WIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0, summarize = TRUE)# An unsummarized set of WIR observations r_WIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0) # A summarized set of of WIR observations r_WIR(n = 5, mu = 2, lambda = 4, stream_length = 20, F_event = F_exp(), F_interim = F_exp(), interval_length = 1, rest_length = 0, summarize = TRUE)
This is a convenience function that allows multiple recording procedures to be applied to a single behavior stream. Results are reported either per behavior stream or as summary statistics, averaged over multiple behavior streams.
reported_observations( BS, data_types = c("C", "M", "E", "P", "W"), interval_length = 1, rest_length = 0, n_aggregate = 1 )reported_observations( BS, data_types = c("C", "M", "E", "P", "W"), interval_length = 1, rest_length = 0, n_aggregate = 1 )
BS |
object of class behavior_stream |
data_types |
list of recording procedures to apply to the behavior stream. See details. |
interval_length |
time length of each interval used to score momentary time recording and interval recording procedures. |
rest_length |
portion of each interval to exclude from observation for interval recording.
See documentation for |
n_aggregate |
number of observations over which to calculate summary statistics. |
The following recording procedures are currently implemented
C - continuous duration recording
M - momentary time recording
E - event counting
P - partial interval recording
W - whole interval recording
If n_aggregate = 1, a data frame with one column per procedure listed in data_types and length equal to the number
of behavior streams in BS. If n_aggregate > 1, a list containing two data frames: one with sample means
and one with sample variances, both taken across n_aggregate behavior streams.
BS <- r_behavior_stream(n = 50, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) reported_observations(BS, interval_length = 10) reported_observations(BS, interval_length = 10, n_aggregate = 5)BS <- r_behavior_stream(n = 50, mu = 3, lambda = 10, F_event = F_exp(), F_interim = F_exp(), stream_length = 100) reported_observations(BS, interval_length = 10) reported_observations(BS, interval_length = 10, n_aggregate = 5)