Package 'circtree' reference manual

Title:	Regression Trees and Forests for Circular Responses
Description:	Infrastructure for fitting distributional trees and forests based on maximum-likelihood estimation of parameters for a circular response, as well as regression methods for a circular response based on maximum-likelihood estimation are provided. For both approaches the von Mises distribution is employed as circular response distribution.
Authors:	Moritz N. Lang [aut, cre] , Lisa Schlosser [aut] , Achim Zeileis [aut]
Maintainer:	Moritz N. Lang <[email protected]>
License:	GPL-2 \| GPL-3
Version:	0.1-0
Built:	2024-11-06 19:13:58 UTC
Source:	https://github.com/r-forge/partykit

Maximum-Likelihood Fitting for a Circular Response

Description

The function circfit carries out maximum-likelihood estimation of parameters for a circular response employing the von Mises distribution. The parameters can be transformed through link functions but do not depend on further covariates (i.e., are constant across observations).

Usage

circfit(y, weights = NULL, start = NULL, start.eta = NULL,
        response_range = NULL,
        vcov = TRUE, type.hessian =  c("checklist", "analytic", "numeric"), 
        method = "L-BFGS-B", estfun = TRUE, optim.control = list(), ...)
circfit(y, weights = NULL, start = NULL, start.eta = NULL,
        response_range = NULL,
        vcov = TRUE, type.hessian =  c("checklist", "analytic", "numeric"), 
        method = "L-BFGS-B", estfun = TRUE, optim.control = list(), ...)

Arguments

`y`	numeric vector of the response
`weights`	optional numeric vector of case weights.
`start`	starting values for the distribution parameters handed over to `optim`
`start.eta`	starting values for the distribution parameters on the link scale handed over to `optim`.
`response_range`	either a logical value indicating whether the response should be transformed to its original range (TRUE) or kept on the interval (-pi,pi] or a two-dimensional vector specifying a range of the circular response.
`vcov`	logical. Specifies whether or not a variance-covariance matrix should be calculated and returned.
`type.hessian`	Can either be 'checklist', 'analytic' or 'numeric' to decide how the hessian matrix should be calculated in the fitting process in `distfit`. For 'checklist' it is checked whether a function 'hdist' is given in the family list. If so, 'type.hessian' is set to 'analytic', otherwise to 'numeric'.
`method`	Optimization which should be applied in `optim`
`estfun`	logical. Should the matrix of observation-wise score contributions (or empirical estimating functions) be returned?
`optim.control`	A list with `optim` control parameters.
`...`	further arguments passed to `optim`.

Details

The function circfit fits the parameter of the von Mises distribution to a circular response variable by applying distfit.

Value

An object of S3 class circfit inheriting from class distfit.

Examples


## example on parameter range:
sdat.par <- circtree_simulate(response_range = c(-pi, pi))
cf.par <- circfit(sdat.par$y)


## example on response range (0, 2pi):
sdat.rad <- circtree_simulate(response_range = c(0, 2*pi))
cf.rad <- circfit(sdat.rad$y)

## example on response range (0, 360):
sdat.deg <- circtree_simulate(response_range = c(0, 360))
cf.deg <- circfit(sdat.deg$y)
## example on parameter range:
sdat.par <- circtree_simulate(response_range = c(-pi, pi))
cf.par <- circfit(sdat.par$y)


## example on response range (0, 2pi):
sdat.rad <- circtree_simulate(response_range = c(0, 2*pi))
cf.rad <- circfit(sdat.rad$y)

## example on response range (0, 360):
sdat.deg <- circtree_simulate(response_range = c(0, 360))
cf.deg <- circfit(sdat.deg$y)

Distributional Regression Forests for a Circular Response

Description

Distributional forests based on maximum-likelihood estimation of parameters for a circular response employing the von Mises distribution.

Usage

circforest(formula, data, response_range = NULL, subset, 
           na.action = na.pass, weights, offset, cluster, strata, 
           control = disttree_control(teststat = "quad", testtype = "Univ", 
           mincriterion = 0, saveinfo = FALSE, minsplit = 20, minbucket = 7, 
           splittry = 2, ...), ntree = 500L, fit.par = FALSE, 
           perturb = list(replace = FALSE, fraction = 0.632),
           mtry = ceiling(sqrt(nvar)), applyfun = NULL, cores = NULL, trace = FALSE, ...)
## S3 method for class 'circforest'
predict(object, newdata = NULL,
        type = c("parameter", "response", "weights", "node"),
        OOB = TRUE, scale = TRUE, response_range = FALSE, ...)
circforest(formula, data, response_range = NULL, subset, 
           na.action = na.pass, weights, offset, cluster, strata, 
           control = disttree_control(teststat = "quad", testtype = "Univ", 
           mincriterion = 0, saveinfo = FALSE, minsplit = 20, minbucket = 7, 
           splittry = 2, ...), ntree = 500L, fit.par = FALSE, 
           perturb = list(replace = FALSE, fraction = 0.632),
           mtry = ceiling(sqrt(nvar)), applyfun = NULL, cores = NULL, trace = FALSE, ...)
## S3 method for class 'circforest'
predict(object, newdata = NULL,
        type = c("parameter", "response", "weights", "node"),
        OOB = TRUE, scale = TRUE, response_range = FALSE, ...)

Arguments

`formula`	a symbolic description of the model to be fit. This should be of type `y ~ x1 + x2` where `y` should be the response variable and `x1` and `x2` are used as partitioning variables.
`data`	an optional data frame containing the variables in the model.
`response_range`	either a logical value indicating whether the response should be transformed to its original range (TRUE) or kept on the interval (-pi,pi] or a two-dimensional vector specifying a range of the circular response.
`subset`	an optional vector specifying a subset of observations to be used in the fitting process.
`na.action`	a function which indicates what should happen when the data contain missing value.
`weights`	optional numeric vector of case weights.
`offset`	an optional vector of offset values.
`cluster`	an optional factor indicating independent clusters. Highly experimental, use at your own risk.
`strata`	an optional factor for stratified sampling.
`control`	a list with control parameters passed to `extree_fit` via `disttree_control` The default values that are not set within the call of `distforest` correspond to those of the default values used by `disttree` from the `disttree` package. `saveinfo = FALSE` leads to less memory hungry representations of trees. Note that arguments `mtry`, `cores` and `applyfun` in `disttree_control` are ignored for `distforest`, because they are already set.
`ntree`	number of trees to grow for the forest.
`fit.par`	logical. if TRUE, fitted and predicted values and predicted parameters are calculated for the learning data (together with loglikelihood)
`perturb`	a list with arguments `replace` and `fraction` determining which type of resampling with `replace = TRUE` referring to the n-out-of-n bootstrap and `replace = FALSE` to sample splitting. `fraction` is the number of observations to draw without replacement.
`mtry`	number of input variables randomly sampled as candidates at each node for random forest like algorithms. Bagging, as special case of a random forest without random input variable sampling, can be performed by setting `mtry` either equal to `Inf` or manually equal to the number of input variables.
`applyfun`	an optional `lapply`-style function with arguments `function(X, FUN, ...)`. It is used for computing the variable selection criterion. The default is to use the basic `lapply` function unless the `cores` argument is specified (see below).
`cores`	numeric. If set to an integer the `applyfun` is set to `mclapply` with the desired number of `cores`.
`trace`	a logical indicating if a progress bar shall be printed while the forest grows.
`object`	an object as returned by `circforest`
`newdata`	an optional data frame containing test data.
`type`	a character string denoting the type of predicted value returned. For `"parameter"` the predicted distributional parameters are returned on the range of (-pi, pi] and for `"response"` the expectation on the range of the response is returned (`response_range`). `"weights"` returns an integer vector of prediction weights. For `type = "node"`, a list of terminal node ids for each of the trees in the forest ist returned.
`OOB`	a logical defining out-of-bag predictions (only if `newdata = NULL`).
`scale`	a logical indicating scaling of the nearest neighbor weights by the sum of weights in the corresponding terminal node of each tree. In the simple regression forest, predicting the conditional mean by nearest neighbor weights will be equivalent to (but slower!) the aggregation of means.
`...`	arguments to be used to form the default `control` argument if it is not supplied directly.

Details

Distributional regression forests for a circular response are an application of model-based recursive partitioning and unbiased recursive partitioning based on the implementation in distforest using the infrastructure of extree_fit.

Value

An object of S3 class circforest inheriting from class distforest.

Examples

#sdat <- circtree_simulate()
#cf <- circforest(y ~ x1 + x2, data = sdat, ntree = 50)
#sdat <- circtree_simulate()
#cf <- circforest(y ~ x1 + x2, data = sdat, ntree = 50)

Circular Regression with Maximum Likelihood Estimation

Description

Fit a regression model for a circular response by maximum likelihood estimation employing the von Mises distribution.

Usage

circmax(formula, data, subset, na.action,
  model = TRUE, y = TRUE, x = FALSE,
  control = circmax_control(...), ...)

circmax_fit(x, y, z = NULL, control)

circmax_control(maxit = 5000, start = NULL, method = "Nelder-Mead",
  solve_kappa = "Newton-Fourier", 
  gradient = FALSE, hessian = TRUE, ...)
circmax(formula, data, subset, na.action,
  model = TRUE, y = TRUE, x = FALSE,
  control = circmax_control(...), ...)

circmax_fit(x, y, z = NULL, control)

circmax_control(maxit = 5000, start = NULL, method = "Nelder-Mead",
  solve_kappa = "Newton-Fourier", 
  gradient = FALSE, hessian = TRUE, ...)

Arguments

`formula`	a formula expression of the form `y ~ x \| z` where `y` is the response and `x` and `z` are regressor variables for the location and the concentration of the von Mises distribution.
`data`	an optional data frame containing the variables occurring in the formulas; y has to be given in radians.
`subset`	an optional vector specifying a subset of observations to be used for fitting.
`na.action`	a function which indicates what should happen when the data contain `NA`s.
`model`	logical. If `TRUE` model frame is included as a component of the returned value.
`x`, `y`	for `circmax`: logical. If `TRUE` the model matrix and response vector used for fitting are returned as components of the returned value. For `circmax_fit`: `x` is a design matrix with regressors for the location and `y` is a vector of observations given in radians.
`z`	a design matrix with regressors for the concentration.
`...`	arguments to be used to form the default `control` argument if it is not supplied directly.
`control`, `maxit`, `start`	a list of control parameters passed to `optim`.
`method`	The `method` to be used for optimization.
`solve_kappa`	Which kappa solver should be used for the starting values for kappa. By default a Newton Fourier is used (`"Newton-Fourier"`). Alternatively, a uniroot provides a safe option (`"Uniroot"`) or code"Banerjee_et_al_2005" provides a quick approximation).
`gradient`	logical. Should gradients be used for optimization? If `TRUE`, the default `method` is `"BFGS"`. Otherwise `method = "Nelder-Mead"` is used.
`hessian`	logical or character. Should a numeric approximation of the (negative) Hessian matrix by `optim` be computed?

Details

circmax fits a regression model for a circular response assuming a von Mises distribution.

circmax_fit is the lower level function where the parameters of the von Mises distribution are fitted by maximum likelihood estimation.

Value

An object of class "circmax".

Examples

## Example 1: Simulated Data:

sdat <- circmax_simulate(n = 1000, beta = c(3, 5, 2), gamma = c(3, 3))

(m1.circmax <- circmax(y ~ x1 + x2 | x3, data = sdat))

## Example 2: Periwinkle Dataset of Fisher and Lee, 1992:
require("circular")
distance <- c(107, 46, 33, 67, 122, 69, 43, 30, 12, 25, 37, 69, 5, 83, 
  68, 38, 21, 1, 71, 60, 71, 71, 57, 53, 38, 70, 7, 48, 7, 21, 27)
directdeg <- c(67, 66, 74, 61, 58, 60, 100, 89, 171, 166, 98, 60, 197, 
  98, 86, 123, 165, 133, 101, 105, 71, 84, 75, 98, 83, 71, 74, 91, 38, 200, 56)
cdirect <- circular(directdeg * 2 * pi/360)
plot(as.numeric(cdirect) ~ distance, ylim = c(0, 4*pi), pch = 20)
points(as.numeric(cdirect) + 2*pi ~ distance, pch = 20)

(m2.circ <- lm.circular(type = "c-l", y = cdirect, x = distance, init = 0.0))
(m2.circmax <- circmax(cdirect ~ distance, data = data.frame(cbind(distance, cdirect))))
## Example 1: Simulated Data:

sdat <- circmax_simulate(n = 1000, beta = c(3, 5, 2), gamma = c(3, 3))

(m1.circmax <- circmax(y ~ x1 + x2 | x3, data = sdat))

## Example 2: Periwinkle Dataset of Fisher and Lee, 1992:
require("circular")
distance <- c(107, 46, 33, 67, 122, 69, 43, 30, 12, 25, 37, 69, 5, 83, 
  68, 38, 21, 1, 71, 60, 71, 71, 57, 53, 38, 70, 7, 48, 7, 21, 27)
directdeg <- c(67, 66, 74, 61, 58, 60, 100, 89, 171, 166, 98, 60, 197, 
  98, 86, 123, 165, 133, 101, 105, 71, 84, 75, 98, 83, 71, 74, 91, 38, 200, 56)
cdirect <- circular(directdeg * 2 * pi/360)
plot(as.numeric(cdirect) ~ distance, ylim = c(0, 4*pi), pch = 20)
points(as.numeric(cdirect) + 2*pi ~ distance, pch = 20)

(m2.circ <- lm.circular(type = "c-l", y = cdirect, x = distance, init = 0.0))
(m2.circmax <- circmax(cdirect ~ distance, data = data.frame(cbind(distance, cdirect))))

Simulated Data Set for `circmax`

Description

This function creates artifical data set for testing the regression models for a circular response by maximum likelihood estimation.

Usage

  circmax_simulate(n = 1000, beta = c(3, 5, 2), gamma = c(3, 3), seed = 111)
circmax_simulate(n = 1000, beta = c(3, 5, 2), gamma = c(3, 3), seed = 111)

Arguments

`n`	The number of Observations.
`beta`	The coefficients for the intercept and the covariates of the location part.
`gamma`	The coefficients for the intercept and the covariates of the concentration part.
`seed`	Sets the 'seed' to a numeric value.

Value

Data frame with simualated covariates and respective response.

Distributional Regression Tree for a Circular Response

Description

Distributional trees based on maximum-likelihood estimation of parameters for a circular response employing the von Mises distribution.

Usage

circtree(formula, data, response_range = NULL, subset, na.action = na.pass,
         weights, offset, cluster, control = disttree_control(...),
         converged = NULL, scores = NULL, doFit = TRUE, ...)
circtree(formula, data, response_range = NULL, subset, na.action = na.pass,
         weights, offset, cluster, control = disttree_control(...),
         converged = NULL, scores = NULL, doFit = TRUE, ...)

Arguments

`formula`	a symbolic description of the model to be fit. This should be of type `y ~ x1 + x2` where `y` should be the response variable and `x1` and `x2` are used as partitioning variables.
`data`	an optional data frame containing the variables in the model.
`response_range`	either a logical value indicating whether the response should be transformed to its original range (TRUE) or kept on the interval (-pi,pi] or a two-dimensional vector specifying a range of the circular response.
`subset`	an optional vector specifying a subset of observations to be used in the fitting process.
`na.action`	a function which indicates what should happen when the data contain missing value.
`weights`	optional numeric vector of case weights.
`offset`	an optional vector of offset values.
`cluster`	an optional factor indicating independent clusters. Highly experimental, use at your own risk.
`control`	control arguments passed to `extree_fit` via `disttree_control`.
`converged`	an optional function for checking user-defined criteria before splits are implemented.
`scores`	an optional named list of scores to be attached to ordered factors.
`doFit`	a logical indicating if the tree shall be grown (TRUE) or not FALSE
`...`	arguments to be used to form the default `control` argument if it is not supplied directly.

Details

Distributional regression trees for a circular response are an application of model-based recursive partitioning and unbiased recursive partitioning based on the implementation in disttree using the infrastructure of extree_fit.

Value

An object of S3 class circtree inheriting from class disttree.

Examples


## example on parameter range:
sdat.par <- circtree_simulate(response_range = c(-pi, pi))
ct.par <- circtree(y ~ x1 + x2, data = sdat.par)
plot(ct.par)

## example on response range (0, 2pi):
sdat.rad <- circtree_simulate(response_range = c(0, 2*pi))
ct.rad <- circtree(y ~ x1 + x2, data = sdat.rad)
## default: type = "response"
plot(ct.rad, tp_args = list(response_range = FALSE))
plot(ct.rad, tp_args = list(response_range = TRUE))
plot(ct.rad, tp_args = list(response_range = c(0, 24)))

## example on response range (0, 360):
sdat.deg <- circtree_simulate(response_range = c(0, 360))
ct.deg <- circtree(y ~ x1 + x2, data = sdat.deg)
plot(ct.deg, tp_args = list(response_range = FALSE))
plot(ct.deg, tp_args = list(response_range = TRUE))
plot(ct.deg, tp_args = list(template = "geographics"))

## example on response range (0, 24):
sdat.hour <- circtree_simulate(response_range = c(0, 24))
ct.hour <- circtree(y ~ x1 + x2, data = sdat.hour, response_range = c(0, 24))
plot(ct.hour, tp_args = list(response_range = FALSE))
plot(ct.hour, tp_args = list(template = "clock24"))
plot(ct.hour, tp_args = list(template = "clock24", 
  circlab = c("no", "mo", "mi", "ev")))
## example on parameter range:
sdat.par <- circtree_simulate(response_range = c(-pi, pi))
ct.par <- circtree(y ~ x1 + x2, data = sdat.par)
plot(ct.par)

## example on response range (0, 2pi):
sdat.rad <- circtree_simulate(response_range = c(0, 2*pi))
ct.rad <- circtree(y ~ x1 + x2, data = sdat.rad)
## default: type = "response"
plot(ct.rad, tp_args = list(response_range = FALSE))
plot(ct.rad, tp_args = list(response_range = TRUE))
plot(ct.rad, tp_args = list(response_range = c(0, 24)))

## example on response range (0, 360):
sdat.deg <- circtree_simulate(response_range = c(0, 360))
ct.deg <- circtree(y ~ x1 + x2, data = sdat.deg)
plot(ct.deg, tp_args = list(response_range = FALSE))
plot(ct.deg, tp_args = list(response_range = TRUE))
plot(ct.deg, tp_args = list(template = "geographics"))

## example on response range (0, 24):
sdat.hour <- circtree_simulate(response_range = c(0, 24))
ct.hour <- circtree(y ~ x1 + x2, data = sdat.hour, response_range = c(0, 24))
plot(ct.hour, tp_args = list(response_range = FALSE))
plot(ct.hour, tp_args = list(template = "clock24"))
plot(ct.hour, tp_args = list(template = "clock24", 
  circlab = c("no", "mo", "mi", "ev")))

Simulated Data Set for `circtree`

Description

This function creates artifical data set for testing the regression trees employing a von Mises distribution.

Usage

  circtree_simulate(n = 1000, mu = c(0, 2, 5), kappa = c(3, 3, 1), 
  response_range = c(0, 2 * pi), seed = 111)
circtree_simulate(n = 1000, mu = c(0, 2, 5), kappa = c(3, 3, 1), 
  response_range = c(0, 2 * pi), seed = 111)

Arguments

`n`	The number of Observations.
`mu`	The distribution parameters for the location part. Currently exactly three parameters necessary.
`kappa`	The distribution parameters for the concentration part. Currently exactly three parameters necessary.
`response_range`	Defines range of simulated response.
`seed`	Sets the 'seed' to a numeric value.

Value

Data frame with simualated covariates and respective response.

Methods for CIRCMAX Objects

Description

Methods for extracting information from fitted circmax objects.

Usage

## S3 method for class 'circmax'
coef(object, model = c("full", "location", "concentration"), ...)
## S3 method for class 'circmax'
terms(x, model = c("location", "concentration", "full"), ...)
## S3 method for class 'circmax'
coef(object, model = c("full", "location", "concentration"), ...)
## S3 method for class 'circmax'
terms(x, model = c("location", "concentration", "full"), ...)

Arguments

`object`, `x`	an object of class `"circmax"`.
`model`	model for which coefficients shall be returned.
`...`	further arguments passed to or from other methods.

Details

In addition to the methods above, a set of standard extractor functions for "circmax" objects is available, including methods to the generic functions print, logLik, and model.frame. Additionally, estfun, vcov provide methods for 'robust' inference.

Von Mises Family 'Dist-List' for `disttree`.

Description

Exported Von Mises Family for implementation in disttree.

Usage

  dist_vonmises(useC = FALSE, ncores = 1)
dist_vonmises(useC = FALSE, ncores = 1)

Arguments

`useC`	logical; if TRUE C routines are used.
`ncores`	Number of cores for parallelization with openMP (No big improvements in terms of running time).

Von Mises Density

Description

Density function for the von Mises distribution with location parameter mu and concentration parameter kappa.

Usage

dvonmises(y, mu, kappa, log = FALSE)
dvonmises(y, mu, kappa, log = FALSE)

Arguments

`y`	vector of observations.
`mu`	vector of location parameters.
`kappa`	vector of concentration parameters.
`log`	logical; if TRUE, probabilities p are given as log(p)

Value

Von Mises Density

Plotting a Regression Tree with a Circular Response (under development).

Description

This function plots regression trees with a circular response based on plot.constparty.

Usage

  ## S3 method for class 'circtree'
plot(x, terminal_panel = node_circular,
    tp_args = list(), tnex = NULL, drop_terminal = NULL, ...)
## S3 method for class 'circtree'
plot(x, terminal_panel = node_circular,
    tp_args = list(), tnex = NULL, drop_terminal = NULL, ...)

Arguments

`x`	Object of class `circtree`.
`terminal_panel`	Do not change.
`tp_args`	Do not change.
`tnex`	Do not change.
`drop_terminal`	Do not change.
`...`	Do not change.

Predicted/Fitted Values for CIRCMAX Fits

Description

Obtains various types of predictions for circmax models.

Usage

## S3 method for class 'circmax'
predict(object, newdata = NULL, type = c("location", "concentration", 
  "parameter"), 
  na.action = na.pass, ...)
## S3 method for class 'circmax'
predict(object, newdata = NULL, type = c("location", "concentration", 
  "parameter"), 
  na.action = na.pass, ...)

Arguments

`object`	an object of class `"circmax"`.
`newdata`	an optional data frame in which to look for variables which to predict.
`type`	type of prediction: `"location"` returns the location of the predicted distribution. `"scale"` returns the scale of the predicted distribution. `"parameter"` returns a data frame with predicted location and scale parameters.
`na.action`	a function which indicates what should happen when the data contain `NA`s. Default is na.pass
`...`	further arguments passed to or from other methods.

Value

For type "location", or "scale" a vector with either the location or the scale of the predicted distribution.

Von Mises Family for `bamlss`.

Description

Exported Von Mises Family for implementation in bamlss.

Usage

  vonmises_bamlss(...)
vonmises_bamlss(...)

Arguments

...

Not used.

Package 'circtree'

Help Index

Maximum-Likelihood Fitting for a Circular Response

Description

Usage

Arguments

Details

Value

See Also

Examples

Distributional Regression Forests for a Circular Response

Description

Usage

Arguments

Details

Value

See Also

Examples

Circular Regression with Maximum Likelihood Estimation

Description

Usage

Arguments

Details

Value

Examples

Simulated Data Set for circmax

Description

Usage

Arguments

Value

Distributional Regression Tree for a Circular Response

Description

Usage

Arguments

Details

Value

See Also

Examples

Simulated Data Set for circtree

Description

Usage

Arguments

Value

Methods for CIRCMAX Objects

Description

Usage

Arguments

Details

See Also

Von Mises Family 'Dist-List' for disttree.

Description

Usage

Arguments

Von Mises Density

Description

Usage

Arguments

Value

Plotting a Regression Tree with a Circular Response (under development).

Description

Usage

Arguments

Predicted/Fitted Values for CIRCMAX Fits

Description

Usage

Arguments

Value

See Also

Von Mises Family for bamlss.

Description

Usage

Arguments

Simulated Data Set for `circmax`

Simulated Data Set for `circtree`

Von Mises Family 'Dist-List' for `disttree`.

Von Mises Family for `bamlss`.