MCGLM Manual

Package ‘mcglm’
April 10, 2018

Type Package
Title Multivariate Covariance Generalized Linear Models
Version 0.4.0
Date 2018-04-10
Author Wagner Hugo Bonat [aut, cre],
Walmes Marques Zeviani [ctb],
Fernando de Pol Mayer [ctb]
Maintainer Wagner Hugo Bonat <[email protected]>
Description Fitting multivariate covariance generalized linear
models (McGLMs) to data. McGLM is a general framework for non-normal
multivariate data analysis, designed to handle multivariate response
variables, along with a wide range of temporal and spatial correlation
structures defined in terms of a covariance link function combined
with a matrix linear predictor involving known matrices.
The models take non-normality into account in the conventional way
by means of a variance function, and the mean structure is modelled
by means of a link function and a linear predictor.
The models are fitted using an efficient Newton scoring algorithm
based on quasi-likelihood and Pearson estimating functions, using
only second-moment assumptions. This provides a unified approach to
a wide variety of different types of response variables and covariance
structures, including multivariate extensions of repeated measures,
time series, longitudinal, spatial and spatio-temporal structures.
The package offers a user-friendly interface for fitting McGLMs
similar to the glm() R function.
See Bonat (2018) <doi:10.18637/jss.v084.i04>, for more information
and examples.
Depends R (>= 3.2.1)
Suggests testthat, plyr, lattice, latticeExtra, knitr, rmarkdown,
MASS, mvtnorm, tweedie, devtools
Imports stats, Matrix, assertthat, graphics
License GPL-3 | file LICENSE
LazyData TRUE
1
2 R topics documented:
URL https://github.com/wbonat/mcglm
BugReports https://github.com/wbonat/mcglm/issues
Encoding UTF-8
VignetteBuilder knitr
RoxygenNote 6.0.1
NeedsCompilation no
R topics documented:
ahs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
anova.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
coef.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
confint.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
ESS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
fitted.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
fit_mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
gof . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
GOSHO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Hunting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
mc_bias_corrected_std . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
mc_car . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
mc_complete_data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
mc_compute_rho . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
mc_conditional_test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
mc_dglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
mc_dist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
mc_id . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
mc_initial_values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
mc_link_function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
mc_ma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
mc_matrix_linear_predictor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
mc_mixed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
mc_ns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
mc_remove_na . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
mc_robust_std . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
mc_rw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
mc_sic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
mc_sic_covariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
mc_twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
mc_variance_function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
NewBorn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
pAIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
pBIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
pKLIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
plogLik . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
ahs 3
plot.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
print.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
residuals.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
RJC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
soil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
soya . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
summary.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
vcov.mcglm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Index 48
ahs Australian Health Survey
Description
The Australian health survey was used by Bonat and Jorgensen (2016) as an example of multivariate
count regression model. The data consists of five count response variables concerning health system
access measures and nine covariates concerning social conditions in Australian for 1987-88.
• sex - Factor with levels male and female.

• age - Respondent’s age in years divided by 100.
• income - Respondent’s annual income in Australian dollars divided by 1000.
• levyplus - Coded factor. If respondent is covered by private health insurance fund for private
patients in public hospital with doctor of choice (1) or otherwise (0).
• freepoor - Coded factor. If respondent is covered by government because low income, recent
immigrant, unemployed (1) or otherwise (0).
• freerepa - Coded factor. If respondent is covered free by government because of old-age or
disability pension, or because invalid veteran or family of deceased veteran (1) or otherwise
(0).
• illnes - Number of illnesses in past 2 weeks, with 5 or illnesses coded as 5.
• actdays - Number of days of reduced activity in the past two weeks due to illness or injury.
• hscore - Respondent’s general health questionnaire score using Goldberg’s method. High
score indicates poor health.
• chcond - Factor with three levels. If respondent has chronic condition(s) and is limited in
activity (limited), or if the respondent has chronic condition(s) but is not limited in activity
(nonlimited) or otherwise (otherwise, reference level).
• Ndoc - Number of consultations with a doctor or specialist (response variable).
• Nndoc - Number of consultations with health professionals (response variable).
• Nadm - Number of admissions to a hospital, psychiatric hospital, nursing or convalescence
home in the past 12 months (response variable).
• Nhosp - Number of nights in a hospital during the most recent admission.
• Nmed - Total number of prescribed and non prescribed medications used in the past two days.
4 anova.mcglm
Usage
data(ahs)
Format
a data.frame with 5190 records and 15 variables.
Source
Deb, P. and Trivedi, P. K. (1997) Demand for medical care by the elderly: A finite mixture approach.
Journal of Applied Econometrics 12(3):313–336.
Bonat, W. H. and Jorgensen, B. (2016) Multivariate covariance generalized linear models. Journal
of Royal Statistical Society - Series C 65:649–675.
Examples
require(mcglm)
data(ahs, package="mcglm")
form1 <- Ndoc ~ income + age
form2 <- Nndoc ~ income + age
Z0 <- mc_id(ahs)
fit.ahs <- mcglm(linear_pred = c(form1, form2),
matrix_pred = list(Z0, Z0), link = c("log","log"),
variance = c("poisson_tweedie","poisson_tweedie"),
data = ahs)
summary(fit.ahs)
anova.mcglm Anova Tables
Description
Performs Wald tests of the significance for the linear predictor components by response variables.
This function is useful for joint hypothesis tests of regression coefficients associated with categorical
covariates with more than two levels. It is not designed for model comparison.
Usage
## S3 method for class 'mcglm'
anova(object, ...)
Arguments
object an object of class mcglm, usually, a result of a call to mcglm() function.
... additional arguments affecting the summary produced. Note that there is no
extra options for mcglm object class.
coef.mcglm 5
Value
A data.frame with Chi-square statistic to test the null hypothesis of a parameter, or a set of pa-
rameters, be zero. Degree of freedom (Df) and p-values. The Wald test based on the observed
covariance matrix of the parameters is used.
Author(s)
Wagner Hugo Bonat, <[email protected]>
Examples
x1 <- seq(-1, 1, l = 100)
x2 <- gl(5, 20)
beta <- c(5, 0, -2, -1, 1, 2)
X <- model.matrix(~ x1 + x2)
set.seed(123)
y <- rnorm(100, mean = X%*%beta, sd = 1)
data = data.frame("y" = y, "x1" = x1, "x2" = x2)
fit.anova <- mcglm(c(y ~ x1 + x2), list(mc_id(data)), data = data)
anova(fit.anova)
coef.mcglm Model Coefficients
Description
Extract model coefficients for objects of mcglm class.
Usage
coef(object, std.error = FALSE, response = c(NA,
1:length(object$beta_names)), type = c("beta", "tau", "power",
"correlation"), ...)
Arguments
object an object of mcglm class.
std.error logical. If TRUE returns the standard errors for the estimates. Default is FALSE.
response a numeric vector specifyng for which response variable the coefficients should
be returned.
type a string vector (can be 1 element length) specifying which coefficients should
be returned.
Options are "beta", "tau", "power", "tau" and "correlation".
6 confint.mcglm
Value
A data.frame with parameters names, estimates, response variable number and parameters type.
Author(s)
confint.mcglm Confidence Intervals for Model Parameters
Description
Computes confidence intervals for parameters in a fitted mcglm model.
Usage

confint(object, parm, level = 0.95, ...)
Arguments
object a fitted mcglm object.

parm specifies for which parameters are to be given confidence intervals, either a vec-
tor of numbers or a vector of strings. If missing, all parameters are considered.
level the nominal confidence level.
... additional arguments affecting the confidence interval produced. Note that there
is no extra options for mcglm object class.
Value
A data.frame with confidence intervals, parameters names, response variable number and param-
eters type.
Author(s)

ESS 7
ESS Generalized Error Sum of Squares
Description
Extract the generalized error sum of squares (ESS) for objects of mcglm class.
Usage
ESS(object, verbose = TRUE)
Arguments
object an object or a list of objects representing a model of mcglm class.
verbose logical. Print or not the ESS value.
Value
Returns the value of the generalized error sum of squares (ESS).
Author(s)
Source
Bonat, W. H. (2018). Multiple Response Variables Regression Models in R: The mcglm Package.
Journal of Statistical Software, 84(4):1–30.
Wang, M. (2014). Generalized Estimating Equations in Longitudinal Data Analysis: A Review and
Recent Developments. Advances in Statistics, 1(1)1–13.
See Also
gof, plogLik, pAIC, pKLIC, GOSHO and RJC.
fitted.mcglm Fitted Values
Description
Extract fitted values for objects of mcglm class.
Usage
fitted(object, ...)
8 fit_mcglm
Arguments
Value
A matrix with fitted values.
Author(s)
fit_mcglm Chaser and Reciprocal Likelihood Algorithms
Description
This function implements the two main algorithms used for fitting multivariate covariance general-
ized linear models. The chaser and the reciprocal likelihood algorithms.
Usage
fit_mcglm(list_initial, list_link, list_variance,
list_covariance, list_X, list_Z, list_offset,
list_Ntrial, list_power_fixed, list_sparse,
y_vec, correct, max_iter, tol, method,
tuning, verbose)
Arguments
list_initial a list of initial values for regression and covariance parameters.
list_link a list specifying the link function names.
Options are: "logit", "probit", "cauchit", "cloglog", "loglog", "identity",
"log", "sqrt", "1/mu^2" and "inverse".
See mc_link_function for details. Default link = "identity".
list_variance a list specifying the variance function names. Options are: "constant", "tweedie",
"poisson_tweedie", "binomialP" and "binomialPQ". See mc_variance_function
for details. Default variance = "constant".
list_covariance
a list of covariance function names. Options are: "identity", "inverse" and
"expm". Default covariance = "identity".
list_X a list of design matrices. See model.matrix for details.
list_Z a list of knowm matrices to compose the matrix linear predictor.
list_offset a list of offset values. Default NULL.
fit_mcglm 9
list_Ntrial a list of number of trials, useful only when analysing binomial data. Default 1.
list_power_fixed
a list of logicals indicating if the power parameters should be estimated or not.
Default power_fixed = TRUE.
list_sparse a list of logicals indicating if the matrices should be set up as sparse matrices.
This argument is useful only when using exponential-matrix covariance link
function. In the other cases the algorithm detects automatically if the matrix
should be sparse or not.
y_vec a vector of the stacked response variables.
correct a logical indicating if the algorithm will use the correction term or not. Default
correct = TRUE.
max_iter maximum number of iterations. Default max_iter = 20.
tol a numeric specyfing the tolerance. Default tol = 1e-04.
method a string specyfing the method used to fit the models ("chaser" or "rc"). Default
method = "chaser".
tuning a numeric value in general close to zero for the rc method and close to 1 for the
chaser method. This argument control the step-length. Default tuning = 1.
verbose a logical if TRUE print the values of the covariance parameters used on each
iteration. Default verbose = FALSE
Value
A list with estimated regression and covariance parameters. Details about the estimation procedures
as iterations, sensitivity, variability are also provided. In general the users do not need to use this
function directly. The mcglm provides GLM interface for fitting mcglm .
Author(s)
Source
See Also
mcglm, mc_matrix_linear_predictor, mc_link_function and

mc_variance_function.
10 gof
gof Measures of Goodness-of-Fit
Description
Extract the pseudo Gaussian log-likelihood (plogLik), pseudo Akaike Information Criterion (pAIC),
pseudo Kullback-Leibler Information Criterion (pKLIC) and pseudo Bayesian Information Crite-
rion (pBIC) for objects of mcglm class.
Usage
gof(object)
Arguments
Value
Returns a data frame containing goodness-of-fit measures.
Author(s)
Source
See Also
plogLik, pAIC, pKLIC and pBIC.

GOSHO 11
GOSHO Gosho Information Criterion
Description
Extract the Gosho Information Criterion (GOSHO) for an object of mcglm class. WARNING: This
function is limited to models with ONE response variable. This function is general useful only for
longitudinal data analysis.
Usage
GOSHO(object, id, verbose = TRUE)
Arguments

id a vector which identifies the clusters or groups. The length and order of id should
be the same as the number of observations. Data are assumed to be sorted so
that observations on a cluster are contiguous rows for all entities in the formula.
verbose logical. Print or not the GOSHO value.
Value
The value of the GOSHO criterion. Note that the function assumes that the data are in the correct
order.
Author(s)
Source
See Also
gof, plogLik, pAIC, pKLIC, ESS and RJC.

12 Hunting
Hunting Hunting in Pico Basile, Bioko Island, Equatorial Guinea.
Description
Case study analysed in Bonat et. al. (2016) concernings on data of animals hunted in the village
of Basile Fang, Bioko Norte Province, Bioko Island, Equatorial Guinea. Monthly number of blue
duikers and other small animals shot or snared was collected for a random sample of 52 commercial
hunters from August 2010 to September 2013. For each animal caught, the species, sex, method of
capture and altitude were documented. The data set has 1216 observations.
• ALT - Factor five levels indicating the Altitude where the animal was caught.
• SEX - Factor two levels Female and Male.
• METHOD - Factor two levels Escopeta and Trampa.
• OT - Monthly number of other small animals hunted.
• BD - Monthly number of blue duikers hunted.
• OFFSET - Monthly number of hunter days.
• HUNTER - Hunter index.
• MONTH - Month index.
• MONTHCALENDAR - Month using calendar numbers (1-January, ..., 12-December).
• YEAR - Year calendar (2010–2013).
• HUNTER.MONTH - Index indicating observations taken at the same HUNTER and MONTH.
Usage
data(Hunting)
Format
Source
Bonat, et. al. (2017). Modelling the covariance structure in marginal multivariate count mod-
els: Hunting in Bioko Island. Journal of Agricultural Biological and Environmental Statistics,
22(4):446–464.
mcglm 13
Examples
library(mcglm)
library(Matrix)
data(Hunting, package="mcglm")
formu <- OT ~ METHOD*ALT + SEX + ALT*poly(MONTH, 4)
Z0 <- mc_id(Hunting)
Z1 <- mc_mixed(~0 + HUNTER.MONTH, data = Hunting)
fit <- mcglm(linear_pred = c(formu), matrix_pred = list(c(Z0, Z1)),
link = c("log"), variance = c("poisson_tweedie"),
power_fixed = c(FALSE),
control_algorithm = list(max_iter = 100),
offset = list(log(Hunting$OFFSET)), data = Hunting)
summary(fit)
anova(fit)
mcglm Fitting Multivariate Covariance Generalized Linear Models
Description
The function mcglm is used to fit multivariate covariance generalized linear models. The models are
specified by a set of lists giving a symbolic description of the linear and matrix linear predictors.
The user can choose between a list of link, variance and covariance functions. The models are fitted
using an estimating function approach, combining quasi-score functions for regression parameters
and Pearson estimating function for covariance parameters. For details see Bonat and Jorgensen
(2016).
Usage
mcglm(linear_pred, matrix_pred, link, variance, covariance,
offset, Ntrial, power_fixed, data, control_initial,
contrasts, control_algorithm)
Arguments
linear_pred a list of formula see formula for details.
matrix_pred a list of known matrices to be used on the matrix linear predictor. For details see
mc_matrix_linear_predictor.
link a list of link functions names. Options are: "logit", "probit", "cauchit",
"cloglog", "loglog", "identity", "log", "sqrt", "1/mu^2" and "inverse".
See mc_link_function for details.
variance a list of variance functions names. Options are: "constant", "tweedie", "poisson_tweedie",
"binomialP" and "binomialPQ".
See mc_variance_function for details.
14 mc_bias_corrected_std
covariance a list of covariance link functions names. Options are: "identity", "inverse"
and exponential-matrix "expm".
offset a list of offset values if any.
Ntrial a list of number of trials on Bernoulli experiments. It is useful only for binomialP
and binomialPQ variance functions.
power_fixed a list of logicals indicating if the values of the power parameter should be esti-
mated or not.
data a data frame.
control_initial
a list of initial values for the fitting algorithm. If no values are supplied automatic
initial values will be provided by the function mc_initial_values.
contrasts extra arguments to passed to model.matrix.
control_algorithm
a list of arguments to be passed for the fitting algorithm. See fit_mcglm for
details.
Value
mcglm returns an object of class ’mcglm’.
Author(s)
Source
See Also
fit_mcglm, mc_link_function and mc_variance_function.
mc_bias_corrected_std Bias-corrected Standard Error for Regression Parameters
Description
Compute bias-corrected standard error for regression parameters in the context of clustered obser-
vations for an object of mcglm class. It is also robust and has improved finite sample properties.
Usage
mc_bias_corrected_std(object, id)
mc_car 15
Arguments
id a vector which identifies the clusters. The length and order of id should be the
same as the number of observations. The data set are assumed to be sorted so
that observations on a cluster are contiguous rows for all entities.
Value
A variance-covariance matrix. Note that the function assumes that the data are in the correct order.
Author(s)
Source
Nuamah, I. F. and Qu, Y. and Aminu, S. B. (1996). A SAS macro for stepwise correlated binary
regression. Computer Methods and Programs in Biomedicine 49, 199–210.
See Also
mc_robust_std.
mc_car Conditional Autoregressive Model Structure
Description
The function mc_car helps to build the components of the matrix linear predictor used for fitting
conditional autoregressive models. This function is used in general for fitting spatial areal data
using the well known conditional autoregressive models (CAR). This function depends on a list of
neighboors, such a list can be constructed, for example using the tri2nb function from the spdep
package based on spatial coordinates. This way to specify the matrix linear predictor can also be
applied for spatial continuous data, as an approximation.
Usage
mc_car(list_neigh, intrinsic = FALSE)
Arguments
list_neigh list of neighboors.
intrinsic logical.
Value
A list of a matrix (intrinsic = TRUE) or two matrices (intrinsic = FALSE).
16 mc_complete_data
Author(s)
Source
See Also
mc_id, mc_compute_rho, mc_conditional_test, mc_dist, mc_ma, mc_rw
and mc_mixed.
mc_complete_data Complete Data Set with NA
Description
The function mc_complete_data completes a data set with NA values for helping to construct the
components of the matrix linear predictor in models that require equal number of observations by
subjects (id).
Usage
mc_complete_data(data, id, index, id.exp)
Arguments
data a data.frame to be completed with NA.
id name of the column (string) containing the subject id.
index name of the column (string) containing the index to be completed.
id.exp how the index is expected to be for all subjects.
Value
A data.frame with the same number of observations by subject.
Author(s)
Source
See Also
mc_dglm, mc_ns, mc_ma and mc_rw.
mc_compute_rho 17
mc_compute_rho Autocorrelation Estimates
Description
Compute autocorrelation estimates based on a fitted model using the mc_car structure. The mcglm
approach fits models using a linear covariance structure, but in general in this parametrization for
spatial models the parameters have no simple interpretation in terms of spatial autocorrelation. The
function mc_compute_rho computes the autocorrelation based on a fitted model.
Usage
mc_compute_rho(object, level = 0.975)
Arguments

level the confidence level required.
Value
Returns estimate, standard error and confidential interval for the spatial autocorrelation parameter.
Author(s)
Source
See Also
mc_car and mc_conditional_test.

18 mc_conditional_test
mc_conditional_test Conditional Hypotheses Tests
Description
Compute conditional hypotheses tests for fitted mcglm model class. When fitting models with extra
power parameters, the standard errors associated with the dispersion parameters can be large. In
that cases, we suggest to conduct conditional hypotheses test instead of the orthodox marginal test
for the dispersion parameters. The function mc_conditional_test offers an ease way to conduct
such conditional test. Furthermore, the function is quite flexible and can be used for any other
conditional hypotheses test.
Usage
mc_conditional_test(object, parameters, test, fixed)
Arguments
object an object representing a model of mcglm class.

parameters which parameters will be included in the conditional test.
test index indicating which parameters will be tested given the values of the other
parameters.
fixed index indicating which parameters should be fixed on the conditional test.
Value
Returns estimates, conditional standard errors and Z-statistics.
Author(s)
Source
mc_dglm 19
mc_dglm Double Generalized Linear Models Structure
Description
The function mc_dglm builds the components of the matrix linear predictor used for fitting double
generalized linear models.
Usage
mc_dglm(formula, id, data)
Arguments
formula a formula spefying the components of the covariance structure.

id name of the column (string) containing the subject index. (If ts is not repeated
measures, use id = 1 for all observations).
data data set.
Value
A list of a diagonal matrices, whose values are given by the covariates assumed to describe the
covariance structure.
Author(s)
Source
See Also
mc_id, mc_dist, mc_ma, mc_rw

and mc_mixed.
20 mc_dist
mc_dist Distance Models Structure
Description
The function mc_dist helps to build the components of the matrix linear predictor using matrices
based on distances. This function is generaly used for the analysis of longitudinal and spatial data.
The idea is to use the inverse of some measure of distance as for example the Euclidean distance
to model the covariance structure within response variables. The model can also use the inverse of
distance squared or high order power.
Usage
mc_dist(id, time, data, method = "euclidean")
Arguments
id name of the column (string) containing the subject index. For spatial data use
the same id for all observations (one unit sample).
time name of the column (string) containing the index indicating the time. For spatial
data use the same index for all observations.
data data set.
method distance measure to be used.
Details
The distance measure must be one of "euclidean", "maximum", "manhattan", "canberra", "binary"
or "minkowski". This function is a customize call of the dist function.
Value
A matrix of dgCMatrix class.
Author(s)
Source
See Also
dist, mc_id, mc_conditional_test, mc_car, mc_ma, mc_rw and mc_mixed.
mc_id 21
Examples
id <- rep(1:2, each = 4)
time <- rep(1:4, 2)
data <- data.frame("id" = id, "time" = time)
mc_dist(id = "id", time = "time", data = data)
mc_dist(id = "id", time = "time", data = data, method = "canberra")
mc_id Independent Model Structure
Description
Builds an identity matrix to be used as a component of the matrix linear predictor. It is in general
the first component of the matrix linear predictor, a kind of intercept matrix.
Usage
mc_id(data)
Arguments
data the data set to be used.
Value
A list of matrix.
Author(s)
Source
See Also
mc_dist, mc_ma, mc_rw and mc_mixed.

22 mc_initial_values
mc_initial_values Automatic Initial Values
Description
This function provides initial values to be used when fitting multivariate covariance generalized
linear models by using the function mcglm. In general the users do not need to use this function,
since it is already employed when setting the argument control_initial = "automatic" in the
mcglm function. However, if the users want to change some of the initial values, this function can
be useful.
Usage
mc_initial_values(linear_pred, matrix_pred, link, variance,
covariance, offset, Ntrial, contrasts, data)
Arguments
linear_pred a list of formula see formula for details.
matrix_pred a list of known matrices to be used on the matrix linear predictor.
See mc_matrix_linear_predictor for details.
link a list of link functions names, see mcglm for details.
variance a list of variance functions names, see mcglm for details.
covariance a list of covariance link functions names, see mcglm for details.
offset a list of offset values if any.
Ntrial a list of the number of trials on Bernoulli experiments. It is useful only for
"binomialP" and "binomialPQ" variance functions.
contrasts list of contrasts to be used in the model.matrix.
data data frame.
Details
To obtain initial values for multivariate covariance generalized linear models the function
mc_initial_values fits a generalized linear model (GLM) using the function glm with the spec-
ified linear predictor and link function for each response variables considering independent obser-
vations. The family argument is always specified as quasi. The link function depends on the
specification of the argument link. The variance function is always specified as "mu" the only ex-
cession appears when using variance = "constant" then the family argument in the glm function
is specified as quasi(link = link, variance = "constant"). The estimated value of the
dispersion parameter from the glm function is used as initial value for the first component of the
matrix linear predictor, for all other components the value zero is used.
For the cases covariance = "inverse" and covariance = "expm" the inverse and the logarithm
of the estimated dispersion parameter is used as initial value for the first component of the matrix
linear predictor. The value of the power parameter is always started at 1. In the cases of multivariate
models the correlation between response variables is always started at 0.
mc_link_function 23
Value
Return a list of initial values to be used while fitting in the mcglm function.
Author(s)
mc_link_function Link Functions
Description
The mc_link_function is a customized call of the make.link function.
Given the name of a link function, it returns a list with two elements. The first element is the
inverse of the link function applied on the linear predictor µ = g −1 (Xβ). The second element is
the derivative of µ with respect to the regression parameters β. It will be useful when computing
the quasi-score function.
Usage
mc_link_function(beta, X, offset, link)
mc_logit(beta, X, offset)
mc_probit(beta, X, offset)
mc_cauchit(beta, X, offset)
mc_cloglog(beta, X, offset)
mc_loglog(beta, X, offset)
mc_identity(beta, X, offset)
mc_log(beta, X, offset)
mc_sqrt(beta, X, offset)
mc_invmu2(beta, X, offset)
mc_inverse(beta, X, offset)
Arguments
beta a numeric vector of regression parameters.
X a design matrix, see model.matrix for details.
24 mc_link_function
offset a numeric vector of offset values. It will be sum up on the linear predictor
as a covariate with known regression parameter equals one (µ = g −1 (Xβ +
of f set)). If no offset is present in the model, set offset = NULL.
link a string specifying the name of the link function. Options are: "logit", "probit",
"cauchit", "cloglog", "loglog", "identity", "log", "sqrt", "1/mu^2"
and inverse. A user defined link function can be used (see Details).
Details
The link function is an important component of the multivariate covariance generalized linear mod-
els, since it links the expectation of the response variable with the covariates. Let β be a (p x 1)
regression parameter vector and X be an (n x p) design matrix. The expected value of the response
variable Y is given by
E(Y ) = g −1 (Xβ),
where g is the link function and η = Xβ is the linear predictor. Let D be a (n x p) matrix whose
entries are given by the derivatives of µ with respect to β. Such a matrix will be required for the
fitting algorithm. The function mc_link_function returns a list where the first element is µ (n x 1)
vector and the second is the D (n x p) matrix. A user defined function can also be used. It must be a
function with arguments beta, X and offset (set to NULL if non needed). The function must return
a length 2 named list with mu and D elements as a vector and a matrix of proper dimensions.
Value
A list with two elements: mu and D (see Details).
Author(s)
See Also
model.matrix, make.link.
Examples
x1 <- seq(-1, 1, l = 5)
X <- model.matrix(~ x1)
mc_link_function(beta = c(1,0.5), X = X,
offset = NULL, link = 'log')
mc_link_function(beta = c(1,0.5), X = X,
offset = rep(10,5), link = 'identity')
mc_ma 25
mc_ma Moving Average Models Structure
Description
The function mc_ma helps to build the components of the matrix linear predictor associated with
moving average models. This function is generaly used for the analysis of longitudinal and times
series data. The user can specify the order of the moving average process.
Usage
mc_ma(id, time, data, order = 1)
Arguments
id name of the column (string) containing the subject index. Note that this structure
was designed to deal with longitudinal data. For times series data use the same
id for all observations (one unit sample).
time name of the column (string) containing the index indicating the time.
data data set.
order order of the moving average process.
Details
This function was designed mainly to deal with longitudinal data, but can also be used for times se-
ries analysis. In that case, the id argument should contain only one index. It pretends a longitudinal
data taken just for one individual or unit sample. This function is a simple call of the bandSparse
function from the Matrix package.
Value
A matrix of dgCMatrix class.
Author(s)
Source
See Also
mc_id, mc_dist, mc_car, mc_rw and mc_mixed.
26 mc_matrix_linear_predictor
Examples
id <- rep(1:2, each = 4)
time <- rep(1:4, 2)
mc_ma(id = "id", time = "time", data = data, order = 1)
mc_ma(id = "id", time = "time", data = data, order = 2)
mc_matrix_linear_predictor
Matrix Linear Predictor
Description
Compute the matrix linear predictor. It is an internal function, however, since the concept of matrix
linear predictor was proposed recently. I decided let this function visible to the interested reader
gets some feeling about how it works.
Usage
mc_matrix_linear_predictor(tau, Z)
Arguments
tau a numeric vector of dispersion parameters.
Z a list of known matrices.
Details
Given a list with a set of known matrices (Z0 , ..., ZD ) the function
mc_matrix_linear_predictor returns U = τ0 Z0 + ... + τD ZD .
Value
A matrix.
Author(s)
Source
mc_mixed 27
See Also
mc_id, mc_dist, mc_ma, mc_rw, mc_mixed and mc_car.
Examples
require(Matrix)
Z0 <- Diagonal(5, 1)
Z1 <- Matrix(rep(1,5)%*%t(rep(1,5)))
Z <- list(Z0, Z1)
mc_matrix_linear_predictor(tau = c(1,0.8), Z = Z)
mc_mixed Mixed Models Structure
Description
The function mc_mixed helps to build the components of the matrix linear predictor associated with
mixed models. It is useful to model the covariance structure as a function of known covariates
in a linear mixed model fashion (Bonat, et. al. 2016). The mc_mixed function was designed to
analyse repeated measures and longitudinal data, where in general the observations are taken at a
fixed number of groups, subjects or unit samples.
Usage
mc_mixed(formula, data)
Arguments
formula a formula model to build the matrix linear predictor. See details.
data data set.
Details
The formula argument should be specified similar to the linear predictor for the mean structure,
however the first component should be 0 and the second component should always indicate the
name of the column containing the subject or unit sample index. It should be a factor. The other
covariates are specified after a slash "\" in the usual way. For example, ~0 + SUBJECT/(x1 + x2)
means that the column SUBJECT contains the subject or unit sample index, while the covariates
that can be continuous or factors are given in the columns x1 and x2. Be careful the parenthesis
after the "\" are mandatory, when including more than one covariate. The special case where only
the SUBJECT effect is requested the formula takes the form ~ 0 + SUBJECT without any extra
covariate. This structure corresponds to the well known compound symmetry structure. By default
the function mc_mixed include all interaction terms, the users can ignore the interactions terms
removing them from the matrix linear predictor.
Value
A list of matrices.
28 mc_ns
Author(s)
Source
22(4):446–464.
See Also
mc_id, mc_conditional_test, mc_dist, mc_ma, mc_rw and mc_car.
Examples
SUBJECT <- gl(2, 6)

x1 <- rep(1:6, 2)
x2 <- rep(gl(2,3),2)
data <- data.frame(SUBJECT, x1 , x2)
# Compound symmetry structure
mc_mixed(~0 + SUBJECT, data = data)
# Compound symmetry + random slope for x1 and interaction or correlation
mc_mixed(~0 + SUBJECT/x1, data = data)
# Compound symmetry + random slope for x1 and x2 plus interactions
mc_mixed(~0 + SUBJECT/(x1 + x2), data = data)
mc_ns Non-structure Model Structure
Description
The function mc_non builds the components of the matrix linear predictor used for fitting non-
structured covariance matrix. In general this model is hard to fit due to the large number of param-
eters.
Usage
mc_ns(id, data, group = NULL, marca = NULL)

mc_remove_na 29
Arguments
id name of the column (string) containing the subject index. Note this structure
was designed to deal with longitudinal data. For times series or spatial data use
the same id for all observations (one unit sample).
data data set.
group name of the column (string) containing a group specific for which the covariance
should change.
marca level (string) of the column group for which the covariance should change.
Value
A list of a n*(n-1)/2 matrices.
Author(s)
Source
See Also
mc_id, mc_dglm, mc_dist, mc_ma, mc_rw
and mc_mixed.
mc_remove_na Remove NA from Matrix Linear Predictor
Description
The function mc_remove_na removes NA from each component of the matrix linear predictor. It is
in general used after the function mc_complete_data.
Usage
mc_remove_na(matrix_pred, cod)
Arguments
matrix_pred a list of known matrices.
cod index indicating the columns should be removed.
Value
A list of matrices.
30 mc_robust_std
Author(s)
Source
See Also
mc_dglm, mc_ns, mc_ma and mc_rw.
mc_robust_std Robust Standard Error for Regression Parameters
Description
Compute robust standard error for regression parameters in the context of clustered observations for
an object of mcglm class.
Usage
mc_robust_std(object, id)
Arguments
id a vector which identifies the clusters or subject indexes. The length and order of
id should be the same as the number of observations. Data are assumed to be
sorted so that observations on a cluster are contiguous rows for all entities in the
formula.
Value
A variance-covariance matrix. Note that the function assumes that the data are in the correct order.
Author(s)
Source
Nuamah, I. F. and Qu, Y. and Aminu, S. B. (1996). A SAS macro for stepwise correlated binary
regression. Computer Methods and Programs in Biomedicine 49, 199–210.
See Also
mc_bias_correct_std.
mc_rw 31
mc_rw Random Walk Models Structure
Description
The function mc_rw builds the components of the matrix linear predictor associated with random
walk models. This function is generaly used for the analysis of longitudinal and times series data.
The user can specify the order of the random walk process.
Usage
mc_rw(id, time, data, order = 1, proper = FALSE)
Arguments
id name of the column (string) containing the subject index. Note that this structure
was designed to deal with longitudinal data. For times series data use the same
id for all observations (one unit sample).
time name of the column (string) containing the index indicating the time.
data data set.
order order of the random walk model.
proper logical.
Value
If proper = FALSE a matrix of dgCMatrix class. If proper = TRUE a list with two matrices of
dgCMatrix class.
Author(s)
Source
See Also
mc_id, mc_dist, mc_car, mc_ma, mc_mixed and mc_compute_rho.

32 mc_sic
Examples
id <- rep(1:2, each = 4)
time <- rep(1:4, 2)
mc_rw(id = "id", time = "time", data = data, order = 1, proper = FALSE)
mc_rw(id = "id", time = "time", data = data, order = 1, proper = TRUE)
mc_rw(id = "id", time = "time", data = data, order = 2, proper = TRUE)
mc_sic Score Information Criterion - Regression
Description
Compute the score information criterion (SIC) for an object of mcglm class. The SIC is useful for
selecting the components of the linear predictor. It can be used to construct an stepwise covariate
selection.
Usage
mc_sic(object, scope, data, response, penalty = 2)
Arguments
scope a vector of covariate names to be tested.
data data set containing all variables involved in the model.
response index indicating for which response variable the SIC should be computed.
penalty penalty term (default = 2).
Value
A data frame containing SIC values, degree of freedom, Tu-statistics and chi-squared reference
values.
Author(s)
Source
22(4):446–464.
mc_sic_covariance 33
See Also
mc_sic_covariance.
Examples
set.seed(123)
x1 <- runif(100, -1, 1)
x2 <- gl(2,50)
beta = c(5, 0, 3)
X <- model.matrix(~ x1 + x2)
y <- rnorm(100, mean = X%*%beta , sd = 1)
data <- data.frame(y, x1, x2)
# Reference model
fit0 <- mcglm(c(y ~ 1), list(mc_id(data)), data = data)
# Computing SIC
mc_sic(fit0, scope = c("x1","x2"), data = data, response = 1)
mc_sic_covariance Score Information Criterion - Covariance
Description
Compute the score information criterion (SIC) for an object of mcglm class. The SIC-covariance
is useful for selecting the components of the matrix linear predictor. It can be used to construct an
stepwise procedure to select the components of the matrix linear predictor.
Usage
mc_sic_covariance(object, scope, idx, data, penalty = 2, response)
Arguments
scope a list of matrices to be tested.
idx indicator of matrices belong to the same effect. It is useful for the case where
more than one matrix represents the same effect.
data data set containing all variables involved in the model.
penalty penalty term (default = 2).
response index indicating for which response variable SIC-covariance should be com-
puted.
Value
A data frame containing SIC-covariance values, degree of freedom, Tu-statistics and chi-squared
reference values for each matrix in the scope argument.
34 mc_twin
Author(s)
Source
22(4):446–464.
See Also
mc_sic.
Examples
set.seed(123)
SUBJECT <- gl(10, 10)
y <- rnorm(100)
data <- data.frame(y, SUBJECT)
Z0 <- mc_id(data)
Z1 <- mc_mixed(~0+SUBJECT, data = data)
# Reference model
fit0 <- mcglm(c(y ~ 1), list(Z0), data = data)
# Testing the effect of the matrix Z1
mc_sic_covariance(fit0, scope = Z1, idx = 1,
data = data, response = 1)
# As expected Tu < Chisq indicating non-significance of Z1 matrix
mc_twin Twin Models Structure
Description
The function mc_twin helps to build the components of the matrix linear predictor associated with
ACDE models for analysis of twin data.
Usage
mc_twin(id, twin.id, type, replicate = NULL, structure, data)
mc_twin_bio(id, twin.id, type, replicate = NULL, structure, data)
mc_twin_full(id, twin.id, type, replicate, formula, data)

mc_twin 35
Arguments
id name of the column (string) containing the twin index. It should be the same
index (number) for both twins.
twin.id name of the column (string) containing the twin index inside the pair. In general
1 for the first twin and 2 for the second twin.
type name of the column (string) containing the indication of the twin as mz or dz.
It should be a factor with only two levels mz and dz. Be sure that the reference
level is mz.
replicate name of the column (string) containing the index for more than one observation
taken at the same twin pair. It is used for example in twin longitudinal studies.
In that case, the replication column should contain the time index.
structure model type options are full, flex, uns, ACE, ADE, AE, CE and E. See example
for details.
data data set.
formula internal.
Value
A list of matrices of dgCMatrix class.
Author(s)
Source
See Also
mc_id, mc_dist, mc_car, mc_rw, mc_ns, mc_dglm and mc_mixed.
Examples
id <- rep(1:5, each = 4)

id.twin <- rep(1:2, 10)
36 mc_variance_function
mc_variance_function Variance Functions
Description
Compute the variance function and its derivatives with respect to regression, dispersion and power
parameters.
Usage
mc_variance_function(mu, power, Ntrial, variance, inverse,
derivative_power, derivative_mu)
mc_power(mu, power, inverse, derivative_power, derivative_mu)
mc_binomialP(mu, power, inverse, Ntrial,

mc_binomialPQ(mu, power, inverse, Ntrial,

Arguments
mu a numeric vector. In general the output from mc_link_function.
power a numeric value (power and binomialP) or a vector (binomialPQ) of the power
parameters.
Ntrial number of trials, useful only when dealing with binomial response variables.
variance a string specifying the name (power, binomialP or binomialPQ) of the vari-
ance function.
inverse logical. Compute the inverse or not.
derivative_power
logical if compute (TRUE) or not (FALSE) the derivatives with respect to the
power parameter.
derivative_mu logical if compute (TRUE) or not (FALSE) the derivative with respect to the mu
parameter.
Details
The function mc_variance_function computes three features related with the variance function.
Depending on the logical arguments, the function returns V 1/2 and its derivatives with respect to the
parameters power and mu, respectivelly. The output is a named list, completely informative about
what the function has been computed. For example, if inverse = FALSE, derivative_power =
TRUE and derivative_mu = TRUE. The output will be a list, with three elements: V_sqrt, D_V_sqrt_power
and D_V_sqrt_mu.
NewBorn 37
Value
A list with from one to four elements depends on the arguments.
Author(s)
Source
of Royal Statistical Society - Series C X(X):XX–XX.
See Also
mc_link_function.
Examples
x1 <- seq(-1, 1, l = 5)
X <- model.matrix(~x1)
mu <- mc_link_function(beta = c(1, 0.5), X = X, offset = NULL,
link = "logit")
mc_variance_function(mu = mu$mu, power = c(2, 1), Ntrial = 1,
variance = "binomialPQ", inverse = FALSE,
derivative_power = TRUE, derivative_mu = TRUE)
NewBorn Respiratory Physiotherapy on Premature Newborns.
Description
The NewBorn dataset consists of a prospective study to assess the effect of respiratory physiother-
apy on the cardiopulmonary function of ventilated preterm newborn infants with birth weight lower
than 1500 g. The data set was collected and kindly made available by the nursing team of the
Waldemar Monastier hospital, Campo Largo, PR, Brazil. The NewBorn dataset was analysed in
Bonat and Jorgensen (2016) as an example of mixed outcomes regression model.
• Sex - Factor two levels Female and Male.

• GA - Gestational age (weeks).
• BW - Birth weight (mm).
• APGAR1M - APGAR index in the first minute of life.
• APGAR5M - APGAR index in the fifth minute of life.
• PRE - Factor, two levels (Premature: YES; NO).
• HD - Factor, two levels (Hansen’s disease, YES; NO).
• SUR - Factor, two levels (Surfactant, YES; NO).
38 NewBorn
• JAU - Factor, two levels (Jaundice, YES; NO).

• PNE - Factor, two levels (Pneumonia, YES; NO).
• PDA - Factor, two levels (Persistence of ductus arteriosus, YES; NO).
• PPI - Factor, two levels (Primary pulmonary infection, YES; NO).
• OTHERS - Factor, two levels (Other diseases, YES; NO).
• DAYS - Age (days).
• AUX - Factor, two levels (Type of respiratory auxiliary, HOOD; OTHERS).
• RR - Respiratory rate (continuous).
• HR - Heart rate (continuous).
• SPO2 - Oxygen saturation (bounded).
• TREAT - Factor, three levels (Respiratory physiotherapy, Evaluation 1; Evaluation 2; Evalua-
tion 3).
• NBI - Newborn index.
• TIME - Days of treatment.
Usage
data(NewBorn)
Format
Source
Examples
library(mcglm)
library(Matrix)
data(NewBorn, package="mcglm")
formu <- SPO2 ~ Sex + APGAR1M + APGAR5M + PRE + HD + SUR
Z0 <- mc_id(NewBorn)
fit <- mcglm(linear_pred = c(formu), matrix_pred = list(Z0),
link = c("logit"), variance = c("binomialP"),
power_fixed = c(TRUE),
data = NewBorn,
control_algorithm = list(verbose = FALSE, tuning = 0.5))
summary(fit)
pAIC 39
pAIC Pseudo Akaike Information Criterion
Description
Extract the pseudo Akaike information criterion (pAIC) for objects of mcglm class.
Usage
pAIC(object, verbose = TRUE)
Arguments
verbose logical. Print or not the pAIC value.
Value
Returns the value of the pseudo Akaike information criterion (pAIC).
Author(s)
Source
See Also
gof, plogLik, ESS, pKLIC, GOSHO and RJC.
pBIC Pseudo Bayesian Information Criterion
Description
Extract the pseudo Bayesian information criterion (pBIC) for objects of mcglm class.
Usage
pBIC(object, verbose = TRUE)
40 pKLIC
Arguments
verbose logical. Print or not the pBIC value.
Value
Returns the value of the pseudo Bayesian information criterion (pBIC).
Author(s)
Source
See Also
gof, plogLik, ESS, pKLIC, GOSHO and RJC.
pKLIC Pseudo Kullback-Leibler Information Criterion
Description
Extract the pseudo Kullback-Leibler information criterion (pKLIC) for objects of mcglm class.
Usage
pKLIC(object, verbose = TRUE)
Arguments
verbose logical. Print or not the pKLIC value.
Value
Returns the value of the pseudo Kullback-Leibler information criterion.
Author(s)
plogLik 41
Source
See Also
gof, plogLik, ESS, pAIC, GOSHO and RJC.
plogLik Gaussian Pseudo-loglikelihood
Description
Extract the Gaussian pseudo-loglikelihood (plogLik) value for objects of mcglm class.
Usage
plogLik(object, verbose = TRUE)
Arguments
verbose logical. Print or not the plogLik value.
Value
Returns the value of the Gaussian pseudo-loglikelihood.
Author(s)
plot.mcglm Residuals and algorithm check plots
Description
Residual and algorithm check analysis for objects of mcglm class.
Usage
plot(x, type = "residuals", ...)
42 print.mcglm
Arguments
x a fitted mcglm object.
type specify which graphical analysis will be performed. Options are: "residuals"
and "algorithm".
... additional arguments affecting the plot produced. Note that there is no extra
options for mcglm object class.
Value
The function plot.mcglm was designed to offer a fast residuals analysis based on the Pearson
residuals. Current version offers a simple Pearson residuals versus fitted values and a quantile plot.
When using algorithm = TRUE the function will plot a summary of the fitting algorithm shows the
trajectory or iterations of the fitting algorithm. The iterations are shown in terms of values of the
model parameters and also the actually value of the quasi-score and Pearson estimating functions.
Hence, a quickly check of the algorithm convergence is obtained.
Author(s)
See Also
residuals and fitted.
print.mcglm Print
Description
The default print method for an object of mcglm class.
Usage
print(x, ...)
Arguments
x fitted model objects of class mcglm as produced by mcglm().
... further arguments passed to or from other methods.
Author(s)
See Also
summary.
residuals.mcglm 43
residuals.mcglm Residuals
Description
Compute residuals for an object of mcglm class.
Usage
residuals(object, type = "raw", ...)
Arguments
type the type of residuals which should be returned. Options are: "raw" (default),
"pearson" and "standardized".
... additional arguments affecting the residuals produced. Note that there is no extra
Value
The function residuals.mcglm returns a matrix of residuals values.
Author(s)
See Also
fitted.
RJC Rotnitzky-Jewell Information Criterion
Description
Compute the Rotnitzky-Jewell information criterion for an object of mcglm class. WARNINGS:
This function is limited to models with ONE response variable.
Usage
RJC(object, id, verbose = TRUE)
44 soil
Arguments
id a vector which identifies the clusters. The length and order of id should be
the same as the number of observations. Data are assumed to be sorted so that
observations on a cluster are contiguous rows for all entities in the formula.
verbose logical. Print or not the RJC value.
Value
The value of the Rotnitzky-Jewell information criterion. Note that the function assumes that the
data are in the correct order.
Author(s)
Source
See Also
gof, plogLik, pAIC, pKLIC, ESS and GOSHO.
soil Soil Chemistry Properties Data
Description
Soil chemistry properties measured on a regular grid with 10 x 25 points spaced by 5 meters.
• COORD.X - X coordinate.
• COORD.Y - Y coordinate.
• SAND - Sand portion of the sample.
• SILT - Silt portion of the sample.
• CLAY - Clay portion of the sample.
• PHWATER - Soil pH at water.
• CA - Calcium content.
• MG - Magnesium content.
• K - Potassio content.
Usage
data(soil)
soya 45
Format
Source
Examples
data(soil, package="mcglm")
neigh <- tri2nb(soil[,1:2])
Z1 <- mc_car(neigh)
# Linear predictor
form.ca <- CA ~ COORD.X*COORD.Y + SAND + SILT + CLAY + PHWATER
fit.ca <- mcglm(linear_pred = c(form.ca), matrix_pred = list(Z1),
link = "log", variance = "tweedie", covariance = "inverse",
power_fixed = FALSE, data = soil,
control_algorith = list(max_iter = 500, tuning = 0.1))
summary(fit.ca)
# Conditional hypothesis test
mc_conditional_test(fit.ca, parameters = c("power11","tau11","tau12"),
test = 2:3, fixed = 1)
# Spatial autocorrelation
mc_compute_rho(fit.ca)
soya Soybeans
Description
Experiment carried out in a vegetation house with soybeans. The experiment has two plants by
plot with three levels of the factor amount of water in the soil (water) and five levels of potassium
fertilization (pot). The plots were arranged in five blocks (block). Three response variables are of
the interest, namely, grain yield, number of seeds and number of viable peas per plant. The data set
has 75 observations of 7 variables.
• pot - Factor five levels of potassium fertilization.

• water - Factor three levels of amount of water in the soil.
• block - Factor five levels.
• grain - Continuous - Grain yield per plant.
• seeds - Count - Number of seeds per plant.
• viablepeas - Binomial - Number of viable peas per plant.
• totalpeas - Binomial - Total number of peas per plant.
46 summary.mcglm
Usage
data(soya)
Format
Source
Examples
library(mcglm)
library(Matrix)
data(soya, package="mcglm")
formu <- grain ~ block + factor(water) * factor(pot)
Z0 <- mc_id(soya)
fit <- mcglm(linear_pred = c(formu), matrix_pred = list(Z0),
data = soya)
anova(fit)
summary.mcglm Summarizing
Description
The default summary method for an object of mcglm class.
Usage
summary(object, verbose = TRUE, print = c("Regression",
"power", "Dispersion", "Correlation"), ...)
Arguments
verbose logical. Print or not the model summary.
print print only part of the model summary, options are Regression, power, Dispersion
and Correlation.
... additional arguments affecting the summary produced. Note the there is no extra
Value
Print a mcglm object.
vcov.mcglm 47
Author(s)
See Also
print.
vcov.mcglm Variance-Covariance Matrix
Description
Returns the variance-covariance matrix for an object of mcglm class.
Usage
vcov(object, ...)
Arguments
Value
A variance-covariance matrix.
Author(s)
Index
∗Topic datasets mc_identity (mc_link_function), 23

ahs, 3 mc_initial_values, 14, 22
Hunting, 12 mc_inverse (mc_link_function), 23
NewBorn, 37 mc_invmu2 (mc_link_function), 23
soil, 44 mc_link_function, 8, 13, 23, 36, 37
soya, 45 mc_log (mc_link_function), 23
mc_logit (mc_link_function), 23
ahs, 3 mc_loglog (mc_link_function), 23
anova.mcglm, 4 mc_ma, 25
mc_matrix_linear_predictor, 13, 22, 26
bandSparse, 25 mc_mixed, 27
mc_ns, 28
coef.mcglm, 5
mc_power (mc_variance_function), 36
confint.mcglm, 6
mc_probit (mc_link_function), 23
mc_remove_na, 29
dist, 20
mc_robust_std, 30
ESS, 7 mc_rw, 31
mc_sic, 32
fit_mcglm, 8, 14 mc_sic_covariance, 33
fitted.mcglm, 7 mc_sqrt (mc_link_function), 23
formula, 13, 22 mc_twin, 34
mc_twin_bio (mc_twin), 34
gof, 10 mc_twin_full (mc_twin), 34
GOSHO, 11 mc_variance_function, 8, 13, 36
mcglm, 9, 13, 22
Hunting, 12 model.matrix, 8, 14, 22–24
make.link, 23, 24 NewBorn, 37

mc_bias_corrected_std, 14
mc_binomialP (mc_variance_function), 36 pAIC, 39
mc_binomialPQ (mc_variance_function), 36 pBIC, 39
mc_car, 15 pKLIC, 40
mc_cauchit (mc_link_function), 23 plogLik, 41
mc_cloglog (mc_link_function), 23 plot.mcglm, 41
mc_complete_data, 16 print.mcglm, 42
mc_compute_rho, 17
mc_conditional_test, 18 residuals.mcglm, 43
mc_dglm, 19 RJC, 43
mc_dist, 20
mc_id, 21 soil, 44
48
INDEX 49
soya, 45
summary.mcglm, 46
vcov.mcglm, 47

MCGLM Manual

Uploaded by

Copyright:

Available Formats

MCGLM Manual

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MCGLM Manual

Uploaded by

Copyright:

Available Formats

Package ‘mcglm’

April 10, 2018

ahs Australian Health Survey

• sex - Factor with levels male and female.

anova.mcglm Anova Tables

coef.mcglm Model Coefficients

Wagner Hugo Bonat, <[email protected]>

confint.mcglm Confidence Intervals for Model Parameters

Computes confidence intervals for parameters in a fitted mcglm model.

## S3 method for class 'mcglm'

object a fitted mcglm object.

Wagner Hugo Bonat, <[email protected]>

ESS Generalized Error Sum of Squares

fitted.mcglm Fitted Values

fit_mcglm Chaser and Reciprocal Likelihood Algorithms

Wagner Hugo Bonat, <[email protected]>

mcglm, mc_matrix_linear_predictor, mc_link_function and

gof Measures of Goodness-of-Fit

object an object or a list of objects representing a model of mcglm class.

Returns a data frame containing goodness-of-fit measures.

Wagner Hugo Bonat, <[email protected]>

plogLik, pAIC, pKLIC and pBIC.

GOSHO Gosho Information Criterion

GOSHO(object, id, verbose = TRUE)

object an object of mcglm class.

Wagner Hugo Bonat, <[email protected]>

gof, plogLik, pAIC, pKLIC, ESS and RJC.

Hunting Hunting in Pico Basile, Bioko Island, Equatorial Guinea.

a data.frame with 1216 records and 11 variables.

mcglm Fitting Multivariate Covariance Generalized Linear Models

mc_bias_corrected_std Bias-corrected Standard Error for Regression Parameters

mc_car Conditional Autoregressive Model Structure

mc_complete_data Complete Data Set with NA

mc_compute_rho Autocorrelation Estimates

mc_compute_rho(object, level = 0.975)

object an object or a list of objects representing a model of mcglm class.

Wagner Hugo Bonat, <[email protected]>

mc_car and mc_conditional_test.

mc_conditional_test Conditional Hypotheses Tests

mc_conditional_test(object, parameters, test, fixed)

object an object representing a model of mcglm class.

Returns estimates, conditional standard errors and Z-statistics.

Wagner Hugo Bonat, <[email protected]>

mc_dglm Double Generalized Linear Models Structure

mc_dglm(formula, id, data)

formula a formula spefying the components of the covariance structure.

Wagner Hugo Bonat, <[email protected]>

mc_id, mc_dist, mc_ma, mc_rw

mc_dist Distance Models Structure

mc_id Independent Model Structure

data the data set to be used.

Wagner Hugo Bonat, <[email protected]>

mc_dist, mc_ma, mc_rw and mc_mixed.

mc_initial_values Automatic Initial Values

mc_link_function Link Functions

A list with two elements: mu and D (see Details).

Wagner Hugo Bonat, <[email protected]>

mc_ma Moving Average Models Structure

mc_mixed Mixed Models Structure