, Simon Schoenbuchner
[ctb] | Title: | Regression by Composition |
|---|---|
| Description: | Flexible statistical modelling using a modular framework for regression, in which groups of transformations are composed together and act on probability distributions. |
| Authors: | Daniel Farewell [aut, cre, cph]
, Simon Schoenbuchner
[ctb] |
| Maintainer: | Daniel Farewell <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.1.0 |
| Built: | 2024-12-07 06:58:54 UTC |
| Source: | CRAN |
Compute Akaike Information Criterion from a regression by composition
## S3 method for class 'RegressionByComposition' AIC(object, ..., k = 2)## S3 method for class 'RegressionByComposition' AIC(object, ..., k = 2)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
ignored |
k |
numeric, the penalty per parameter to be used; ‘k = 2’ is the classical AIC. |
Append a flow to a CompositeFamily object
append_flow(family, flow)append_flow(family, flow)
family |
a CompositeFamily object |
flow |
a Flow object |
a new CompositeFamily object
append_flow(Normal(0, 1), Translate) Reduce(append_flow, list(Scale, Translate), init = Normal(0, 1))append_flow(Normal(0, 1), Translate) Reduce(append_flow, list(Scale, Translate), init = Normal(0, 1))
Bernoulli distribution as a CompositeFamily
Bernoulli(prob = 0.5)Bernoulli(prob = 0.5)
prob |
the probability of a success |
a new BinaryFamily object
dist <- Bernoulli() dist$probability()dist <- Bernoulli() dist$probability()
Extract regression coefficients from a regression by composition
## S3 method for class 'RegressionByComposition' coef(object, ...)## S3 method for class 'RegressionByComposition' coef(object, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
ignored |
Compute fitted values from a regression by composition
## S3 method for class 'RegressionByComposition' fitted(object, ...)## S3 method for class 'RegressionByComposition' fitted(object, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
further arguments passed to the R6 method $fitted() associated with the model's CompositeFamily |
Extract log-likelihood from a regression by composition
## S3 method for class 'RegressionByComposition' logLik(object, ...)## S3 method for class 'RegressionByComposition' logLik(object, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
ignored |
Lognormal distribution as a CompositeFamily
LogNormal(meanlog = 0, sdlog = 1)LogNormal(meanlog = 0, sdlog = 1)
meanlog |
the mean of the logarithm |
sdlog |
the standard deviation of the logarithm |
a new ContinuousFamily object
dist <- LogNormal() log(dist$quantile(0.95))dist <- LogNormal() log(dist$quantile(0.95))
Moebius flow
MoebiusMoebius
An object of class Flow (inherits from R6) of length 6.
Normal distribution as a CompositeFamily
Normal(mean = 0, sd = 1)Normal(mean = 0, sd = 1)
mean |
the mean |
sd |
the standard deviation |
a new ContinuousFamily object
dist <- Normal() dist$quantile(0.95)dist <- Normal() dist$quantile(0.95)
Power flow
PowerPower
An object of class Flow (inherits from R6) of length 6.
Compute predicted values from a regression by composition
## S3 method for class 'RegressionByComposition' predict(object, newdata, ...)## S3 method for class 'RegressionByComposition' predict(object, newdata, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
newdata |
data.frame containing new data |
... |
further arguments passed to the R6 method $fitted() associated with the model's CompositeFamily |
Fit a regression by composition model
rbc(formula, init, flows, family, data, par, hessian = TRUE)rbc(formula, init, flows, family, data, par, hessian = TRUE)
formula |
a formula object, with model components separated by '|' |
init |
the initial distribution |
flows |
a list of flows |
family |
(optional) an object of class 'CompositeFamily'; if supplied, 'init' and 'flows' are ignored |
data |
a data frame |
par |
a vector of starting values |
hessian |
logical; use Hessian matrix in model fitting? |
an rbc object
## Annette Dobson (1990) ## "An Introduction to Generalized Linear Models". ## Page 9: Plant Weight Data. ctl <- c(4.17, 5.58, 5.18, 6.11, 4.50, 4.61, 5.17, 4.53, 5.33, 5.14) trt <- c(4.81, 4.17, 4.41, 3.59, 5.87, 3.83, 6.03, 4.89, 4.32, 4.69) dobson <- data.frame( weight = c(ctl, trt), group = gl(2, 10, 20, labels = c("Ctl", "Trt")) ) dobson_fit <- rbc(weight ~ 1 | 1 + group, init = Normal(0, 1), flows = list(Scale, Translate), data = dobson ) starr_fit <- rbc( height ~ 1 | 0 + I((280 + age)^(-1)) | 1 | 1, init = LogNormal(), flows = list(Power, Moebius, Scale, Translate), data = subset(starr, id %in% unique(id)[1:10]) )## Annette Dobson (1990) ## "An Introduction to Generalized Linear Models". ## Page 9: Plant Weight Data. ctl <- c(4.17, 5.58, 5.18, 6.11, 4.50, 4.61, 5.17, 4.53, 5.33, 5.14) trt <- c(4.81, 4.17, 4.41, 3.59, 5.87, 3.83, 6.03, 4.89, 4.32, 4.69) dobson <- data.frame( weight = c(ctl, trt), group = gl(2, 10, 20, labels = c("Ctl", "Trt")) ) dobson_fit <- rbc(weight ~ 1 | 1 + group, init = Normal(0, 1), flows = list(Scale, Translate), data = dobson ) starr_fit <- rbc( height ~ 1 | 0 + I((280 + age)^(-1)) | 1 | 1, init = LogNormal(), flows = list(Power, Moebius, Scale, Translate), data = subset(starr, id %in% unique(id)[1:10]) )
Compute 'residuals' from a regression by composition
## S3 method for class 'RegressionByComposition' residuals(object, ...)## S3 method for class 'RegressionByComposition' residuals(object, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
ignored |
a vector of probabilities of the same length as the data
Scale flow
ScaleScale
An object of class Flow (inherits from R6) of length 6.
ScaleOdds flow
ScaleOddsScaleOdds
An object of class Flow (inherits from R6) of length 6.
ScaleRisk0 flow
ScaleRisk0ScaleRisk0
An object of class Flow (inherits from R6) of length 6.
ScaleRisk1 flow
ScaleRisk1ScaleRisk1
An object of class Flow (inherits from R6) of length 6.
Growth from birth to 3 years in healthy babies in the US
starrstarr
starrA data frame with 104,798 rows and 5 columns:
Anonymized identifier
Sex of baby
Age of baby, in days
Jittered height of baby, in cm
Jittered weight of baby, in kg
...
doi:10.1186/s12874-024-02145-1
Summary of a regression by composition
## S3 method for class 'RegressionByComposition' summary(object, compact = FALSE, ...)## S3 method for class 'RegressionByComposition' summary(object, compact = FALSE, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
compact |
logical; should coefficients from all flows be compressed into a single matrix? |
... |
ignored |
Translate flow
TranslateTranslate
An object of class Flow (inherits from R6) of length 6.
TranslateRisk1 flow
TranslateRisk1TranslateRisk1
An object of class Flow (inherits from R6) of length 6.
Extract variance-covariance matrix from a regression by composition
## S3 method for class 'RegressionByComposition' vcov(object, ...)## S3 method for class 'RegressionByComposition' vcov(object, ...)
object |
a RegressionByComposition object; usually the result of a call to rbc() |
... |
ignored |