Package 'mcount'

Promoting Adherence for Life (PAL) study dataset

Description

A dataset from the Promoting Adherence for Life (PAL) study, as reported in Huh, Flaherty, and Simoni (2012) and Simoni et al. (2009). The PAL study was a 2×2 factorial trial that evaluated peer support and text reminders for promoting HIV antiretroviral adherence.

Usage

dat.pal

Format

A data frame with variables including:

id

Participant identifier.

tx

A factor variable specifying the intervention group with the following categories: "CTRL" = Control (Standard of Care), "PEER" = Peer support only, "TXTS" = Text messages only, and "BOTH" = Peer support plus text messages.

doses0

Number of doses taken in the past 3 days, at baseline.

doses3

Number of doses taken in the past 3 days, at post-intervention.

totaldoses

Number of total prescribed doses.

Source

PAL study

References

Huh, D., Flaherty, B. P., & Simoni, J. M. (2012). Optimizing the analysis of adherence interventions using logistic generalized estimating equations. AIDS and Behavior, 16(2), 422–431. doi:10.1007/s10461-011-9955-5

Simoni, J. M., Huh, D., Frick, P. A., Pearson, C. R., Andrasik, M. P., Dunbar, P. J., & Hooton, T. M. (2009). Peer support and pager messaging to promote antiretroviral modifying therapy in Seattle: A randomized controlled trial. Journal of Acquired Immune Deficiency Syndromes, 52(4), 465–473. doi:10.1097/QAI.0b013e3181b9300c

---

Personalized Feedback Intervention (PFI) example dataset

Description

A dataset from White, Mun, and Morgan (2008), which is also described in Mun et al. (2015, 2022).

Usage

dat.pfi

Format

A data frame with 194 rows and 5 variables:

m0

Number of standard alcohol drinks consumed at baseline.

int_PF

Indicator for receiving personalized feedback intervention (PFI): 1 = received PFI, 0 = did not receive PFI.

year_new

Indicator for first-year college student status: 1 = first-year student, 0 = otherwise.

race_new

Indicator for race: 1 = White, 0 = non-White.

y

Number of standard alcohol drinks consumed at post-intervention; the response variable.

Source

White et al. (2008)

References

White, H. R., Mun, E.-Y., & Morgan, T. J. (2008). Do brief personalized feedback interventions work for mandated students or is it just getting caught that works? Psychology of Addictive Behaviors, 22(1), 107–116. doi:10.1037/0893-164X.22.1.107

Mun, E. Y., De La Torre, J., Atkins, D. C., White, H. R., Ray, A. E., Kim, S. Y., ... & The Project INTEGRATE Team. (2015). Project INTEGRATE: An integrative study of brief alcohol interventions for college students. Psychology of Addictive Behaviors, 29(1), 34–48. doi:10.1037/adb0000047

Mun, E.-Y., Zhou, Z., Huh, D., Tan, L., Li, D., Tanner-Smith, E. E., Walters, S. T., & Larimer, M. E. (2022). Brief alcohol interventions are effective through six months: Findings from marginalized zero-inflated Poisson and negative binomial models in a two-step IPD meta-analysis. Prevention Science. doi:10.1007/s11121-022-01420-1

---

Estimating Marginalized Zero-inflated Poisson Models

Description

Fit marginalized zero-inflated Poisson models for zero-inflated count data via maximum likelihood estimation.

Usage

mzip(formula, data)

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. A typical formula has the form response ~ terms where response is the count response vector and terms is a series of terms that predict response. For example, formula = y ~ x1 + x2 + x3. Do not write intercept in the formula; intercept will be automatically added in model fitting.

data

a data frame containing variables in the model.

Details

Function returns an object of class "mle2" from bbmle R package. Apply summary function to the resulting object from the function to obtain more estimation information.

Value

mzip returns an object of class "mzip". The fitted model object contains the estimated coefficients and other information about the fitted model.

The function summary (i.e., summary.mzip) can be used to obtain or print a summary of the results.

Suffix _zero corresponds to the parameters associated with the structrual zero rate part of a model.

Suffix _mean corresponds to the parameters associated with the overall mean, which evaluate the effects of covariates on the overall mean.

References

Long, D. L., Preisser, J. S., Herring, A. H., & Golin, C. E. (2014). A marginalized zero‐inflated Poisson regression model with overall exposure effects. Statistics in Medicine, 33(29), 5151-5165.

Examples

head(dat.pfi)

#Fit a marginalized zero-inflated Poisson model
res = mzip(formula = y ~ m0 + int_PF + year_new + race_new, data = dat.pfi)

#Obtain estimation results
summary(res)

---

Estimating Marginalized Zero- and N-inflated Binomial (MZNIB) models

Description

Fit MZNIB regression models for fractional count data with boundary inflation at zero and the upper limit with bootstrapped standard errors.

Usage

mznib(
  formula,
  data,
  N_i,
  initial_value = "auto",
  initial_value_alpha = "auto",
  R = 1000,
  ncpus = 1,
  optim_method = "L-BFGS-B",
  parallel = "no",
  seed = NULL,
  test_method = "perc"
)

Arguments

formula	A model formula.
data	A data frame containing variables in the model.
N_i	The maximum value that i-th subject can take (vector).
initial_value	Starting values for estimating the coefficients of the zero- and N-inflated binomial (ZNIB) model, which is used as the "working model" from which the marginalized coefficients are derived. The marginalized coefficients are derived from these initial working estimates from the ZINB. This can be specified manually (i.e., a vector whose length is the number of regression coefficients specified in "formula" (+1 for the intercept) and multiplied by 3). The starting values can also be estimated automatically using GLM by specifying "auto", which we recommend.
initial_value_alpha	Starting values for estimating the marginalized coefficients, with length equal to the number of regression coefficients specified in "formula" +1 for the intercept. The starting values can also be estimated automatically using GLM (="auto"), which we recommend.
R	The number of bootstrapped replications for estimating standard errors and confidence intervals. A higher number of replications improves precision.
ncpus	The number of CPUs to use (>1 for parallel processing)
optim_method	The maximum-likelihood optimizer used to estimate the initial ZINB estimates from which the marginalized MZNIB estimates are derived. We recommend specifying "L-BFGS-B".
parallel	The type of parallel processing used, if any. Options include "multicore" (MacOS and Unix-based systems), "snow" (Windows systems), or "no" for single-core processing.
seed	A numeric seed number so that the results are reproducible. The number of CPUs specified in the ncpus argument must also be kept the same for exact reproduction of results.
test_method	The method used to estimate confidence intervals and test statistical significance of the marginalized coefficients. The default is "perc" which uses the percentile-based confidence intervals and corresponding P-values. Robust, normal approximation confidence intervals and P-values can be specified using "MCD", which uses the minimum covariance determinant method.

Value

An object of class "mznib", which contains the fitted MZNIB model and has the following components:

coef

A data frame with the coefficient estimates and associated standard errors, confidence intervals, and p-values.

alpha_estimates

A matrix with the coefficient estimates from each bootstrap replication.

convergence

A logical value indicating whether the model fit for the original data converged.

bootobj

The bootstrap object containing the bootstrap results.

call

The original function call.

formula

The original regression formula.

R

A scalar indicating the number of bootstrap replications.

test_method

A character string indicating the method used to estimate the confidence intervals and test statistical significance.

The function summary (i.e., summary.mznib) can be used to obtain or print a summary of the results.

Examples

data("dat.pal", package = "mcount")

# Prepare covariates
dat.pal$doses0z <- as.numeric(scale(dat.pal$doses0))
dat.pal$tx_peer <- as.integer(dat.pal$tx == "PEER")
dat.pal$tx_txts <- as.integer(dat.pal$tx == "TXTS")
dat.pal$tx_both <- as.integer(dat.pal$tx == "BOTH")

# Fit the MZNIB model
fit <- mznib(
  doses3 ~ doses0z + tx_peer + tx_txts + tx_both,
  data = dat.pal,
  N_i = dat.pal$totaldoses,
  initial_value = "auto",
  initial_value_alpha = "auto",
  R = 150,
  ncpus = 1,
  parallel = "no"
)

summary(fit)

---

Print method for mzip objects

Description

Print method for mzip objects

Usage

## S3 method for class 'mzip'
print(x, ...)

Arguments

x	A fitted mzip object.
...	Additional arguments passed to print().

---

Print method for summary.mznib objects

Description

Print method for summary.mznib objects

Usage

## S3 method for class 'summary.mznib'
print(x, ...)

Arguments

x	A summary.mznib object.
...	Additional arguments (not used).

---

Summary method for mzip objects

Description

Summary method for mzip objects

Usage

## S3 method for class 'mzip'
summary(object, ...)

Arguments

object	A fitted mzip object.
...	Additional arguments passed to bbmle::summary().

Value

A summary object from bbmle::summary().

---

Summary method for mznib objects

Description

Summary method for mznib objects

Usage

## S3 method for class 'mznib'
summary(object, ...)

Arguments

object	A fitted mznib object.
...	Additional arguments (not used).

Value

An object of class "summary.mznib".

---