yunyishen/car-lasso

Conditional Auto-Regressive LASSO in R

car-lasso's Issues

Use CARlasso function with matrices X and Y rather than formula

For higher dimension cases the formula can be too long to approach, we should have CARlasso allowing matrix input.

Function generate igraph objects

This can be done by users themselves but can be better done as a function since we generated the network object during plot methods anyway.

warning and errors: matrix is not symmetric, arguments imply differing number of rows: 11, 9

I am getting the following warnings and errors.

CARlasso(Didymellaceae + Cladosporium ramotenellum+Didymella sp.+ Didymosphaeriaceae + Didymellaceae.1 + Gibberella intricans+Colletotrichum spaethianum+Alternaria alternata+Coniothyrium sp. + all_others~CO2 , data = forlasso,link = "logit", adaptive = TRUE, n_iter = 5000, n_burn_in = 1000, thin_by = 10)

warning: chol(): given matrix is not symmetric

and then with plot

Error in data.frame(id = c(paste0("resp", 1:n_resp), paste0("pred", 1:n_pred)), : arguments imply differing number of rows: 11, 9

Errors:decomposition failed and matrix is singular or not positive definite

Two errors that I've found when running the package are that:

1. Error: chol(): decomposition failed

CARlasso(Ga0485157_metabat1.059+Ga0485167_maxbin.109+ Ga0485162_maxbin.089+Ga0485161_maxbin.110+Ga0485161_maxbin.075+Ga0485157_metabat1.036+Ga0485165_metabat2_ours.012_sub+Ga0485169_maxbin.201_sub+Ga0485172_maxbin.081_sub+Ga0485168_maxbin.153+ Ga0485172_metabat2_ours.083+ Ga0485162_maxbin.023+ Ga0485160_maxbin.092+ Ga0485158_metabat2_jgi.024+ Ga0485162_metabat1.001+**Ga0485163_metabat1.131**~ depth+wtemp+sp_cond+chlor_rfu+phyco_rfu+fdom_rfu+turb_fnu+do_sat+do_raw+ph, data = data_use, link="log",adaptive = TRUE, n_iter = 5000, n_burn_in = 1000, thin_by = 10)

It is only related to the last column of the responses(Ga0485163_metabat1.131). when removing this last one, the error is gone.

2. Error: inv_sympd(): matrix is singular or not positive definite

CARlasso(Ga0485157_metabat1.059+Ga0485167_maxbin.109+ Ga0485162_maxbin.089+Ga0485161_maxbin.110+Ga0485161_maxbin.075+Ga0485157_metabat1.036+Ga0485165_metabat2_ours.012_sub+Ga0485169_maxbin.201_sub+Ga0485172_maxbin.081_sub+Ga0485168_maxbin.153+ Ga0485172_metabat2_ours.083+ Ga0485162_maxbin.023+ Ga0485160_maxbin.092+ Ga0485158_metabat2_jgi.024+ Ga0485162_metabat1.001+**Ga0485157_metabat2_jgi.016**~ depth+wtemp+sp_cond+chlor_rfu+phyco_rfu+fdom_rfu+turb_fnu+do_sat+do_raw+ph, data = data_use, link="log",adaptive = TRUE, n_iter = 5000, n_burn_in = 1000, thin_by = 10)

Also related to the last column of the responses(Ga0485157_metabat2_jgi.016). when removing this last one, the error is gone.

Down below is a small dataset to reproduce the error:
selected_otus.csv

Down below are the reproducible code

library(CARlasso)
data_use <-read.csv("selected_otus.csv")
otu_res <- CARlasso(Ga0485157_metabat1.059+Ga0485167_maxbin.109+ Ga0485162_maxbin.089+Ga0485161_maxbin.110+Ga0485161_maxbin.075+Ga0485157_metabat1.036+Ga0485165_metabat2_ours.012_sub+Ga0485169_maxbin.201_sub+Ga0485172_maxbin.081_sub+Ga0485168_maxbin.153+ Ga0485172_metabat2_ours.083+ Ga0485162_maxbin.023+ Ga0485160_maxbin.092+ Ga0485158_metabat2_jgi.024+ Ga0485162_metabat1.001+Ga0485163_metabat1.131~ depth+wtemp+sp_cond+chlor_rfu+phyco_rfu+fdom_rfu+turb_fnu+do_sat+do_raw+ph, data = data_use, link="log",adaptive = TRUE, n_iter = 5000, n_burn_in = 1000, thin_by = 10)
otu_res <- horseshoe(otu_res)
plot(otu_res)

bGlasso malfunctions with probit link

I've tried to run bGlasso with the same binary dataset as CARlasso (sans predictors). It runs fine with CARlasso and it appears to also be running fine with bGlasso when I use the count dataset and log link instead. The matrix is 122 rows by 25 columns

carNull <- bGlasso(data=phdat[,1:25], link="probit", n_iter=2000, n_burn_in = 1000, thin_by=10)

Algorithm start...
progress:
0% 10 20 30 40 50 60 70 80 90 100%
[----|----|----|----|----|----|----|----|----|----|
**************************************************|
Error in Intercept_Graphical_LASSO_hir_Cpp(y, 3, n_iter, n_burn_in, thin_by, :
mvnrnd(): given covariance matrix is not symmetric positive semi-definite