Description Usage Arguments Details Value References Examples
Sparsity Oriented Importance Learning (SOIL) provides a new variable importance measure for high dimensional linear regression and logistic regression from a sparse penalization perspective, by taking into account the variable selection uncertainty via the use of a sensible model weighting. The package is an implementation of Ye, C., Yang, Y., and Yang, Y. (2017+) DOI: <doi:10.1080/01621459.2017.1377080>.
1 2 3 4 5 6 |
x |
Matrix of predictors. |
y |
Response variable. |
n_train |
Size of training set when the weight function is |
no_rep |
Number of replications when the weight function is |
n_train_bound |
When computing the weights using |
n_bound |
When computing the weights using |
psi |
A positive number to control the improvement of the prior weight. The default value is 1. |
family |
Choose the family for GLM models. So far |
method |
Users can choose |
candidate_models |
Only available when |
weight_type |
Options for computing weights for SOIL measure. Users can choose among |
prior |
Whether to use prior in the weighting function. The default is |
reduce_bias |
If the binomial model is used, occasionally the algorithm might has convergence issue when the problem of so-called complete separation or quasi-complete separation happens. Users can set |
See the paper provided in Reference section.
A "SOIL" object is retured. The components are:
importance |
SOIL importance values for each variable. |
weight |
The weight for each candidate model. |
candidate_models_cleaned |
Cleaned candidate models: the duplicated candidate models are cleaned; When computing SOIL weights using AIC and BIC, the models with more than n-2 variables are removed (n is the number of observaitons); When computing SOIL weights using ARM, the models with more than n_train-2 variables are removed (n_train is the number of training observations). |
Ye, C., Yang, Y., and Yang, Y. (2017+). "Sparsity Oriented Importance Learning for High-dimensional Linear Regression". Journal of the American Statistical Association. (Accepted) DOI: 10.1080/01621459.2017.1377080
BugReport: https://github.com/emeryyi/SOIL
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 | # REGRESSION CASE
# generate simulation data
n <- 50
p <- 8
beta <- c(3,1.5,0,0,2,0,0,0)
b0 <- 1
x <- matrix(rnorm(n*p,0,1),nrow=n,ncol=p)
e <- rnorm(n)
y <- x %*% beta + b0 + e
# compute SOIL using ARM with prior
v_ARM <- SOIL(x, y, family = "gaussian",
weight_type = "ARM", prior = TRUE)
# compute SOIL using BIC
v_BIC <- SOIL(x, y, family = "gaussian", weight_type = "BIC")
# compute SOIL using AIC
v_AIC <- SOIL(x, y, family = "gaussian",
weight_type = "AIC", prior = TRUE)
# user supplied candidate models
candidate_models = rbind(c(0,0,0,0,0,0,0,1),
c(0,1,0,0,0,0,0,1), c(0,1,1,1,0,0,0,1),
c(0,1,1,0,0,0,0,1), c(1,1,0,1,1,0,0,0),
c(1,1,0,0,1,0,0,0))
v1_BIC <- SOIL(x, y,
psi=1,
family = "gaussian",
method = "customize",
candidate_models = candidate_models,
weight_type = "BIC", prior = TRUE)
# CLASSIFICATION CASE
# generate simulation data
n = 300
p = 8
b <- c(1,1,1,-3*sqrt(2)/2)
x=matrix(rnorm(n*p, mean=0, sd=1), n, p)
feta=x[, 1:4]%*%b
fprob=exp(feta)/(1+exp(feta))
y=rbinom(n, 1, fprob)
# compute SOIL for model_check using BIC with prior
b_BIC <- SOIL(x, y, family = "binomial", weight_type = "BIC")
candidate_models =
rbind(c(0,0,0,0,0,0,0,1),
c(0,1,0,0,0,0,0,1),
c(1,1,1,1,0,0,0,0),
c(0,1,1,0,0,0,0,1),
c(1,1,0,1,1,0,0,0),
c(1,1,0,0,1,0,0,0),
c(0,0,0,0,0,0,0,0),
c(1,1,1,1,1,0,0,0))
# compute SOIL for model_check using AIC
# user supplied candidate models
b_AIC <- SOIL(x, y, family = "binomial",
method = "customize", candidate_models = candidate_models,
weight_type = "AIC")
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.