Create an ABC-RF object: a classification random forest from a reference table towards performing an ABC model choice

Description

abcrf constructs a random forest from a reference table towards performing an ABC model choice. Basically, the reference table (i.e. the dataset that will be treated with the present package) includes a column with the index of the models to be compared and additional columns corresponding to the values of the simulated summary statistics.

Usage

## S3 method for class 'formula'
abcrf(formula, data, group=list(), lda=TRUE, ntree=500, sampsize=min(1e5, nrow(data)),
paral=FALSE, ncores= if(paral) max(detectCores()-1,1) else 1, ...)

Arguments

Value

An object of class abcrf, which is a list with the following components:

References

Pudlo P., Marin J.-M., Estoup A., Cornuet J.-M., Gautier M. and Robert, C. P. (2016) Reliable ABC model choice via random forests Bioinformatics doi:10.1093/bioinformatics/btv684

Estoup A., Raynal L., Verdu P. and Marin J.-M. (2018) Model choice using Approximate Bayesian Computation and Random Forests: analyses based on model grouping to make inferences about the genetic history of Pygmy human populations Jounal de la Société Française de Statistique http://journal-sfds.fr/article/view/709

See Also

plot.abcrf, predict.abcrf, err.abcrf, ranger

Examples

data(snp)
modindex <- snp$modindex[1:500]
sumsta <- snp$sumsta[1:500,]
data1 <- data.frame(modindex, sumsta)
model.rf1 <- abcrf(modindex~., data = data1, ntree=100)
model.rf1
model.rf2 <- abcrf(modindex~., data = data1, group = list(c("1","2"),"3"), ntree=100)
model.rf2

Predict posterior covariance between two parameters for new data using two reg-ABC-RF objects

Description

Using two reg-ABC-RF objects constructed on the same reference table for two different response variables, this function predicts the posterior covariance between those two response variables, given a new dataset of summaries.

Usage

## S3 method for class 'regAbcrf'
covRegAbcrf(regForest1, regForest2, obs, training1, training2,
ntree=500, mtry=max(floor((dim(training1)[2]-1)/3), 1),
sampsize=min(1e5, dim(training1)[1]), paral = FALSE, 
ncores = if(paral) max(detectCores()-1,1) else 1, paral.predict = FALSE,
ncores.predict = if(paral.predict) max(detectCores()-1,1) else 1, ... )

Arguments

Value

covRegAbcrf returns predicted posterior covariances between response variables of two reg-ABC-RF objects, for a new data set.

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

regAbcrf, predict.regAbcrf, err.regAbcrf, plot.regAbcrf, ranger, densityPlot

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
N1 <- snp$param$N1[modindex == "3"]
N1 <- N1[1:500]
data3 <- data.frame(N1, sumsta) 
model.rf.N1 <- regAbcrf(N1~., data3, ntree=100)
data(snp.obs)
covRegAbcrf(model.rf.r, model.rf.N1, snp.obs, 
            data2, data3, ntree=100)

Plot the posterior density given a new summary statistic

Description

Given a reg-ABC-RF object and a new value of the summary statistics, densityPlot gives the corresponding posterior density plot of the parameter, as well as the prior (in grey).

Usage

## S3 method for class 'regAbcrf'
densityPlot(object, obs, training,
add=TRUE, main="Posterior density", log="", xlim=NULL,
ylim=NULL, xlab=NULL, ylab=NULL, paral=FALSE,
ncores= if(paral) max(detectCores()-1,1) else 1, ...)

Arguments

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

regAbcrf, predict.regAbcrf, err.regAbcrf, covRegAbcrf, ranger, plot.regAbcrf

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
data(snp.obs)
densityPlot(model.rf.r, snp.obs, data2, ylab="density", main = "Posterior density of r")

Calculate and plot for different numbers of tree, the out-of-bag errors associated with an ABC-RF object

Description

err.abcrf returns out-of-bag errors and plot them.

Usage

err.abcrf(object, training, paral=FALSE,
ncores= if(paral) max(detectCores()-1,1) else 1)

Arguments

Value

A matrix with 2 columns: the number of trees and the out-of-bag errors. Errors are computed from 40 trees to the total number.

References

Pudlo P., Marin J.-M., Estoup A., Cornuet J.-M., Gautier M. and Robert, C. P. (2016) Reliable ABC model choice via random forests Bioinformatics doi:10.1093/bioinformatics/btv684

See Also

abcrf, predict.abcrf, plot.abcrf

Examples

data(snp)
modindex <- snp$modindex[1:500]
sumsta <- snp$sumsta[1:500,]
data1 <- data.frame(modindex, sumsta)
model.rf <- abcrf(modindex~., data1, ntree=100)
err.rf <- err.abcrf(model.rf, data1)

Calculate and plot for different numbers of tree, the out-of-bag mean squared errors associated with a REG-ABC-RF object

Description

err.regAbcrf returns out-of-bag mean squared errors and plot them.

Usage

err.regAbcrf(object, training, paral=FALSE, 
ncores= if(paral) max(detectCores()-1,1) else 1, what="mean")

Arguments

Value

A matrix with 2 columns: the number of trees and the out-of-bag mean squared errors. NAs might be returned if the number of trees is too low. Errors are computed from 40 trees to the total number.

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

regAbcrf, predict.regAbcrf, plot.regAbcrf, densityPlot, covRegAbcrf, ranger

Examples

  data(snp)
  modindex <- snp$modindex
  sumsta <- snp$sumsta[modindex == "3",]
  r <- snp$param$r[modindex == "3"]
  r <- r[1:500]
  sumsta <- sumsta[1:500,]
  data2 <- data.frame(r, sumsta)
  model.rf.r <- regAbcrf(r~., data2, ntree=100)
  err.regAbcrf(model.rf.r, data2)

Plot of an ABC-RF object

Description

plot.abcrf provides both a variable importance plot of a model choice ABC-RF object and the projection of the reference table on the LDA axes.

Usage

## S3 method for class 'abcrf'
plot(x, training, obs=NULL, n.var=20, pdf=FALSE, xlim=NULL, ...)

Arguments

Note

The graph of the reference table projected on the LD axes is shown only if LD axes has been added to the set of summary statistics in the call of abcrf.

References

Pudlo P., Marin J.-M., Estoup A., Cornuet J.-M., Gautier M. and Robert, C. P. (2016) Reliable ABC model choice via random forests Bioinformatics doi:10.1093/bioinformatics/btv684

See Also

abcrf, predict.abcrf, err.abcrf, variableImpPlot

Examples

data(snp)
modindex <- snp$modindex[1:500]
sumsta <- snp$sumsta[1:500,]
data1 <- data.frame(modindex, sumsta)
model.rf <- abcrf(modindex~., data1, ntree=100)
plot(model.rf, data1)
data(snp.obs)
plot(model.rf, data1, obs=snp.obs[1,])

Plot of a reg-ABC-RF object

Description

plot.regAbcrf provides a variable importance plot used to construct the reg-ABC-RF object, as measured by ranger with the argument importance='impurity'.

Usage

## S3 method for class 'regAbcrf'
plot(x, n.var=min(30, length(x$model.rf$variable.importance)), xlim=NULL, main=NULL, ...)

Arguments

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:bioinformatics/bty867

See Also

regAbcrf, predict.regAbcrf, err.regAbcrf, covRegAbcrf, ranger, densityPlot

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
plot(model.rf.r)

Predict and evaluate the posterior probability of the MAP for new data using an ABC-RF object

Description

Based on an ABC-RF object this function predicts the best model for new data and evaluate the posterior probability of the MAP.

Usage

## S3 method for class 'abcrf'
predict(object, obs, training, ntree = 1000, 
sampsize = min(1e5,  object$model.rf$num.samples ), paral = FALSE,
ncores = if(paral) max(detectCores()-1,1) else 1, paral.predict = FALSE, 
ncores.predict = if(paral.predict) max(detectCores()-1,1) else 1, ...)

Arguments

Value

An object of class abcrfpredict, which is a list with the following components:

References

Pudlo P., Marin J.-M., Estoup A., Cornuet J.-M., Gautier M. and Robert, C. P. (2016) Reliable ABC model choice via random forests Bioinformatics doi:10.1093/bioinformatics/btv684

See Also

abcrf, plot.abcrf, err.abcrf

Examples

data(snp)
modindex <- snp$modindex[1:500]
sumsta <- snp$sumsta[1:500,]
data1 <- data.frame(modindex, sumsta)
model.rf <- abcrf(modindex~., data1, ntree=100)
data(snp.obs)
predict(model.rf, snp.obs, data1, ntree=100)

Predict posterior expectation, median, variance and quantiles given a new dataset using a reg-ABC-RF object

Description

Based on a reg-ABC-RF object this function predicts the posterior expectation, median, variance, quantiles for the corresponding parameter given new dataset. Somes posterior errors can be computed at an higher computational price.

Usage

## S3 method for class 'regAbcrf'
predict(object, obs, training, quantiles=c(0.025,0.975),
paral = FALSE, ncores = if(paral) max(detectCores()-1,1) else 1, rf.weights = FALSE,
post.err.med = FALSE, ...)

Arguments

Value

An object of class regAbcrfpredict, which is a list with the following components:

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

regAbcrf, predictOOB, plot.regAbcrf, err.regAbcrf, covRegAbcrf, ranger, densityPlot

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
data(snp.obs)
predict(model.rf.r, snp.obs, data2)

Predict out-of-bag posterior expectation, median, variance, quantiles and error measures using a reg-ABC-RF object

Description

Based on a reg-ABC-RF object this function predicts the out-of-bag posterior expectation, median, variance, quantiles, mean squared errors, normalized mean absolute errors, credible interval and coverage, for the corresponding parameter using the out-of-bag observations of the training data set.

Mean squared errors and normalized mean absolute errors are computed both with mean and median of the response variable.

Memory allocation issues might be encountered when the size of the training data set is large.

Usage

## S3 method for class 'regAbcrf'
predictOOB(object, training, quantiles=c(0.025,0.975), paral = FALSE,
ncores = if(paral) max(detectCores()-1,1) else 1,...)

Arguments

Value

An object of class regAbcrfOOBpredict, which is a list with the following components:

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

regAbcrf, predict.regAbcrf, plot.regAbcrf, err.regAbcrf, covRegAbcrf, ranger, densityPlot

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
res <- predictOOB(model.rf.r, data2) 

Read a reference table simulated from DIYABC

Description

readRefTable reads a reference table simulated from DIYABC thanks to a .bin and a .txt file, respectively containing the reference table and its header.

Usage

readRefTable(filename = "reftable.bin", header = "header.txt", N = 0)

Arguments

Value

A list with the following components:

References

Cornuet J.-M., Pudlo P., Veyssier J., Dehne-Garcia A., Gautier M., Leblois R., Marin J.-M. and Estoup A. (2014) DIYABC v2.0: a software to make Approximate Bayesian Computation inferences about population history using Single Nucleotide Polymorphism, DNA sequence and microsatellite data Bioinformatics doi:10.1093/bioinformatics/btn514

Create a reg-ABC-RF object: a regression random forest from a reference table aimed out predicting posterior expectation, variance and quantiles for a parameter

Description

regAbcrf constructs a regression random forest from a reference table towards predicting posterior expectations, variances and quantiles of a parameter.

Usage

## S3 method for class 'formula'
regAbcrf(formula, data, ntree=500,
mtry=max(floor((dim(data)[2]-1)/3), 1), sampsize=min(1e5, nrow(data)),
paral=FALSE, ncores=if(paral) max(detectCores()-1,1) else 1, ...)

Arguments

Value

An object of class regAbcrf, which is a list with the following components:

References

Raynal L., Marin J.-M. Pudlo P., Ribatet M., Robert C. P. and Estoup, A. (2019) ABC random forests for Bayesian parameter inference Bioinformatics doi:10.1093/bioinformatics/bty867

See Also

plot.regAbcrf, err.regAbcrf, predict.regAbcrf, covRegAbcrf, ranger, densityPlot, predictOOB.

Examples

data(snp)
modindex <- snp$modindex
sumsta <- snp$sumsta[modindex == "3",]
r <- snp$param$r[modindex == "3"]
r <- r[1:500]
sumsta <- sumsta[1:500,]
data2 <- data.frame(r, sumsta)
model.rf.r <- regAbcrf(r~., data2, ntree=100)
model.rf.r

A simulated example in population genetics

Description

The simulated example of population genetics with SNP loci used in Pudlo et al. (2016): snp contains the reference table on which to perform ABC model choice, it also contains the simulated parameters to perform regression random forest. snp.obs contains two pseudo-observed data sets. The first one (favorable) should be easily allocated to a model, while that is not the case for the second one (unfavorable).

Usage

data(snp)
data(snp.obs)

Format

snp is a list containing an ABC reference table of 10,000 simulations from a Bayesian prior predictive model (see Pudlo et al., 2016, for a description of the model choice issue). The first element, named modindex is a factor containing the model indices, the second element, param, is a data frame with seven simulated parameters. The last element of this list, named sumsta, contains the reference table on which to perform ABC model choice and parameter estimation.

snp.obs is a data frame containing the summary statistics of two pseudo-observed data sets.

Source

Pudlo, P., Marin, J.-M., Estoup, A., Cornuet, J.-M., Gautier, M. and Robert, C.P. (2016) Reliable ABC model choice via random forests Bioinformatics doi:10.1093/bioinformatics/btv684

Examples

data(snp)
data(snp.obs)

Variable importance plot from a random forest

Description

variableImpPlot provides a dotchart of variable importance as measured by ranger with the argument importance='impurity'.

Usage

variableImpPlot(object, 
n.var=min(30, length(object$model.rf$variable.importance)),
xlim=NULL, main=NULL)

Arguments

Value

Invisibly, the importance of the variables that were plotted.

See Also

abcrf, plot.abcrf, plot.regAbcrf

Examples

data(snp)
modindex <- snp$modindex[1:500]
sumsta <- snp$sumsta[1:500,]
data1 <- data.frame(modindex, sumsta)
model.rf <- abcrf(modindex~., data1, ntree=100)
variableImpPlot(model.rf)