cotram (version 0.6-0)

mcotram: Multivariate Count Conditional Transformation Models

Description

A proof-of-concept implementation of multivariate conditional transformation models for count data.

Usage

mcotram(..., formula = ~ 1, data, conditional = FALSE, theta = NULL,
        fixed = NULL, scale = FALSE, optim = mltoptim(hessian = TRUE), 
        M = 1000, dofit = TRUE, domargins = TRUE)

Value

An object of class mmlt with coef and predict methods.

Arguments

...

marginal count transformation models, one for each response

formula

a model formula describing a model for the dependency structure via the lambda parameters. The default is set to ~ 1 for constant lambdas.

data

a data.frame.

conditional

logical; parameters are defined conditionally (only possible when all models are probit models). This is the default as described by Klein et al. (2022). If FALSE, parameters can be directly interpreted marginally, this is explained in Section 2.6 by Klein et al. (2022). Using conditional = FALSE with probit-only models gives the same likelihood but different parameter estimates.

theta

an optional vector of starting values.

fixed

an optional named numeric vector of predefined parameter values.

scale

a logical indicating if (internal) scaling shall be applied to the model coefficients.

optim

a list of optimisers as returned by mltoptim

M

number of Halton sequences used to approximate the log-likelihood in lpmvnorm.

dofit

logical; parameters are fitted by default, otherwise a list with log-likelihood and score function is returned.

domargins

logical; all model parameters are fitted by default, including the parameters of marginal models.

Details

The function implements multivariate count conditional transformation models. The response is assumed to be a vector of counts.

References

Luisa Barbani, Roland Brandl, Torsten Hothorn (2022), Multi-species Count Transformation Models, tools:::Rd_expr_doi("10.48550/arXiv.2201.13095").

Nadja Klein, Torsten Hothorn, Luisa Barbanti, Thomas Kneib (2020), Multivariate Conditional Transformation Models. Scandinavian Journal of Statistics, tools:::Rd_expr_doi("10.1111/sjos.12501").

Examples

library("cotram")
data("spiders", package = "cotram")

### for illustration only
OR <- 1      ### order of transformation function
             ### OR = 1 means log-linear, use OR ~ 6
M <- 100     ### number of Halton sequences, seem sufficient here
fastopt <- mltoptim(abstol = 1e-3, reltol = 1e-3)
             ### faster convergence

## fit conditional marginal count transformation models
## one for each species
m_PF <- cotram(Pardosa_ferruginea ~ Elevation + Canopy_openess, 
               data = spiders, method = "probit", order = OR)
m_HL <- cotram(Harpactea_lepida ~ Elevation + Canopy_openess,
               data = spiders, method = "probit", order = OR)
m_CC <- cotram(Callobius_claustrarius ~ Elevation + Canopy_openess,
               data = spiders, method = "probit", order = OR)
m_CT <- cotram(Coelotes_terrestris ~ Elevation + Canopy_openess,
               data = spiders, method = "probit", order = OR)
m_PL <- cotram(Pardosa_lugubris ~ Elevation + Canopy_openess,
               data = spiders, method = "probit", order = OR)
m_PR <- cotram(Pardosa_riparia ~ Elevation + Canopy_openess,
               data = spiders, method = "probit", order = OR)

### fit dependence parameters
mm <- mcotram(m_PF, m_HL, m_CC, m_CT, m_PL, m_PR, data = spiders,
              M = M, scale = TRUE, optim = fastopt)
logLik(mm)

### Kendall's tau: Dependence of species after accounting
### for elevation and canopy openess in marginal models
coef(mm, type = "Kendall")

# \donttest{

### regress dependencies on elevation and canopy openess
mmc <- mcotram(m_PF, m_HL, m_CC, m_CT, m_PL, m_PR, data = spiders, 
               formula = ~ Elevation + Canopy_openess, M = M, 
               scale = TRUE, optim = fastopt)
logLik(mmc)

### weak evidence for such effects
pchisq(2 * (logLik(mmc) - logLik(mm)), df = 30, lower.tail = FALSE)

### plot Kendall's tau for different elevations / openess levels
nd <- expand.grid(Elevation = 80:120 * 10, Canopy_openess = 1:10 * 10)
KD <- Lower_tri(coef(mmc, newdata = nd, type = "Kendall"))
f <- factor(rownames(KD))
nd <- cbind(f = rep(f, nrow(nd)), nd[rep(1:nrow(nd), each = nlevels(f)),])
nd$KD <- c(KD)

if (require("lattice"))
contourplot(KD ~ Elevation + Canopy_openess | f, data = nd, 
            cuts = 18, xlab = "Elevation", ylab = "Canopy openess")

### for example:
### => constant negative dependence of Pardosa_lugubris and Coelotes_terrestris
### => weak dependence of Harpactea_lepida and Pardosa_ferruginea
###    for low elevations, negative dependence increasing with elevation

# }