| Type: | Package |
| Title: | Transformation-Free Linear Regression for Compositional Outcomes and Predictors |
| Version: | 0.1.2 |
| Maintainer: | Jacob Fiksel <jfiksel@gmail.com> |
| Description: | Implements the expectation-maximization (EM) algorithm as described in Fiksel et al. (2021) <doi:10.1111/biom.13465> for transformation-free linear regression for compositional outcomes and predictors. |
| License: | GPL-2 |
| Imports: | SQUAREM (≥ 2020.3), future, future.apply |
| Encoding: | UTF-8 |
| URL: | https://github.com/jfiksel/codalm |
| BugReports: | https://github.com/jfiksel/codalm/issues |
| RoxygenNote: | 7.1.1 |
| Suggests: | knitr, gtools, remotes, testthat, rmarkdown |
| VignetteBuilder: | knitr |
| Depends: | R (≥ 2.10) |
| LazyData: | true |
| NeedsCompilation: | no |
| Packaged: | 2021-07-26 17:25:48 UTC; jfiksel |
| Author: | Jacob Fiksel 
[aut, cre],
Abhirup Datta [aut] |
| Repository: | CRAN |
| Date/Publication: | 2021-07-26 18:40:02 UTC |
Transformation-free Linear Regression for Compositional Outcomes and Predictors
Description
Implements the expectation-maximization (EM) algorithm as described in Fiksel et al. (2020) for transformation-free linear regression for compositional outcomes and predictors.
Usage
codalm(y, x, accelerate = TRUE)
Arguments
y |
A matrix of compositional outcomes. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
x |
A matrix of compositional predictors. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
accelerate |
A logical variable, indicating whether or not to use the Squarem algorithm for acceleration of the EM algorithm. Default is TRUE. |
Value
A D_s x D_r compositional coefficient matrix, where
D_s and D_r are the dimensions of the compositional predictor
and outcome, respectively
References
https://onlinelibrary.wiley.com/doi/full/10.1111/biom.13465
Examples
data("educFM")
father <- as.matrix(educFM[,2:4])
y <- father / rowSums(father)
mother <- as.matrix(educFM[,5:7] )
x <- mother/rowSums(mother)
codalm(y, x)
Bootstrap Confidence Intervals Linear Regression for Compositional Outcomes and Predictors
Description
Implements percentile based bootstrapping to estimate the confidence intervals for the regression coefficients when doing linear regression for compositional outcomes and predictors
Usage
codalm_ci(
y,
x,
accelerate = TRUE,
nboot = 500,
conf = 0.95,
parallel = FALSE,
ncpus = NULL,
strategy = NULL,
init.seed = 123
)
Arguments
y |
A matrix of compositional outcomes. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
x |
A matrix of compositional predictors. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
accelerate |
A logical variable, indicating whether or not to use the Squarem algorithm for acceleration of the EM algorithm. Default is TRUE |
nboot |
The number of bootstrap repetitions to use. Default is 500 |
conf |
A scalar between 0 and 1 containing the confidence level of the required intervals. Default is .95. |
parallel |
A logical variable, indicating whether or not to use a parallel operation for computing the permutation statistics |
ncpus |
Optional argument. When provided, is an integer giving the number of clusters to be used in parallelization. Defaults to the number of cores, minus 1. |
strategy |
Optional argument. When provided, this will be the evaluation function
(or name of it) to use for parallel computation (if parallel = TRUE). Otherwise,
if parallel = TRUE, then this will default to multisession. See |
init.seed |
The initial seed for the permutations. Default is 123. |
Value
A list, with ci_L and ci_U, giving the lower and upper bounds
of each element of the B matrix
Examples
data("educFM")
father <- as.matrix(educFM[,2:4])
y <- father / rowSums(father)
mother <- as.matrix(educFM[,5:7] )
x <- mother/rowSums(mother)
codalm_ci(y, x, nboot = 50, conf = .95)
Permutation Test for Linear Independence Between Compositional Outcomes and Predictors
Description
Implements the loss function based permutation test as described in Fiksel et al. (2020) for a test of linear independence between compositional outcomes and predictors.
Usage
codalm_indep_test(
y,
x,
nperms = 500,
accelerate = TRUE,
parallel = FALSE,
ncpus = NULL,
strategy = NULL,
init.seed = 123
)
Arguments
y |
A matrix of compositional outcomes. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
x |
A matrix of compositional predictors. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
nperms |
The number of permutations. Default is 500. |
accelerate |
A logical variable, indicating whether or not to use the Squarem algorithm for acceleration of the EM algorithm. Default is TRUE. |
parallel |
A logical variable, indicating whether or not to use a parallel operation for computing the permutation statistics |
ncpus |
Optional argument. When provided, is an integer giving the number of clusters to be used in parallelization. Defaults to the number of cores, minus 1. |
strategy |
Optional argument. When provided, this will be the evaluation function
(or name of it) to use for parallel computation (if parallel = TRUE). Otherwise,
if parallel = TRUE, then this will default to multisession. See |
init.seed |
The initial seed for the permutations. Default is 123. |
Value
The p-value for the independence test
Examples
require(gtools)
x <- rdirichlet(100, c(1, 1, 1))
y <- rdirichlet(100, c(1, 1, 1))
codalm_indep_test(y, x)
data("educFM")
father <- as.matrix(educFM[,2:4])
y <- father / rowSums(father)
mother <- as.matrix(educFM[,5:7] )
x <- mother/rowSums(mother)
codalm_indep_test(y, x)
education level of father (F) and mother (M)
Description
Education level of father (F) and mother (M) in percentages of low (l), medium (m), and high (h) of 31 countries in Europe.
Usage
educFM
Format
A data frame with 31 observations (rows) and 7 columns (country and education level)
Details
countrycommunity codeF.lpercentage of females with low edcuation levelF.mpercentage of females with medium edcuation levelF.hpercentage of females with high edcuation levelF.lpercentage of males with low edcuation levelF.mpercentage of males with medium edcuation levelF.hpercentage of males with high edcuation level
Source
from robCompositions R package, https://cran.r-project.org/src/contrib/Archive/robCompositions/robCompositions_2.3.0.tar.gz, which used the dataEurostat, https://ec.europa.eu/eurostat/
Prediction for Transformation-free Linear Regression for Compositional Outcomes and Predictors
Description
Obtains compositional predictions for new compositional covariates using an established codalm model.
Usage
predict_codalm(object, newx)
Arguments
object |
A |
newx |
A matrix of compositional predictors. Each row is an observation, and must sum to 1. If any rows do not sum to 1, they will be renormalized |
Value
A D_s x D_r compositional coefficient matrix, where
D_s and D_r are the dimensions of the compositional predictor
and outcome, respectively
Examples
data("educFM")
father <- as.matrix(educFM[,2:4])
y <- father / rowSums(father)
mother <- as.matrix(educFM[,5:7] )
x <- mother/rowSums(mother)
codalm_model <- codalm(y[1:20,], x[1:20,])
predict_codalm(codalm_model, x[-(1:20),])