From 9ae24c7e6980b141b16a85d18eb11257a902e4ac Mon Sep 17 00:00:00 2001
From: Jack Leary <jrleary@live.unc.edu>
Date: Tue, 4 Jun 2024 14:53:47 -0400
Subject: [PATCH 1/2] updated imports in DESCRIPTION

---
 DESCRIPTION | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/DESCRIPTION b/DESCRIPTION
index fc4089e..7279414 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -9,7 +9,7 @@ Description: scLANE uses truncated power basis spline models to build flexible,
              Downstream analysis functionalities include model comparison, dynamic gene clustering, smoothed counts generation, gene set enrichment testing, & visualization. 
 License: MIT + file LICENSE
 Encoding: UTF-8
-RoxygenNote: 7.2.3
+RoxygenNote: 7.3.1
 Depends: 
   glm2, 
   magrittr, 
@@ -17,6 +17,7 @@ Depends:
 Imports: 
     geeM,
     MASS,
+    mpath, 
     dplyr,
     stats,
     utils, 

From d322f6118be64b5a40952235686d4ba3daaaa5ca Mon Sep 17 00:00:00 2001
From: Jack Leary <jrleary@live.unc.edu>
Date: Wed, 5 Jun 2024 10:53:31 -0400
Subject: [PATCH 2/2] tweaked GLMM LASSO functionality

---
 R/fitGLMM.R    | 39 ++++++++++++++++++++++++++++++++++++---
 man/fitGLMM.Rd |  6 ++++--
 2 files changed, 40 insertions(+), 5 deletions(-)

diff --git a/R/fitGLMM.R b/R/fitGLMM.R
index 8473967..2d09720 100644
--- a/R/fitGLMM.R
+++ b/R/fitGLMM.R
@@ -16,6 +16,7 @@
 #' @param M.glm The number of possible basis functions to use in the calls to \code{\link{marge2}} when choosing basis functions adaptively.
 #' @param return.basis (Optional) Whether the basis model matrix (denoted \code{B_final}) should be returned as part of the \code{marge} model object. Defaults to FALSE.
 #' @param return.GCV (Optional) Whether the final GCV value should be returned as part of the \code{marge} model object. Defaults to FALSE.
+#' param vebose (Optional) Should intermediate output be printed to the console? Defaults to FALSE.
 #' @return An object of class \code{marge} containing the fitted model & other optional quantities of interest (basis function matrix, GCV, etc.).
 #' @seealso \code{\link[glmmTMB]{glmmTMB}}
 #' @seealso \code{\link{testDynamic}}
@@ -38,7 +39,8 @@ fitGLMM <- function(X_pred = NULL,
                     approx.knot = TRUE,
                     M.glm = 3,
                     return.basis = FALSE,
-                    return.GCV = FALSE) {
+                    return.GCV = FALSE,
+                    verbose = FALSE) {
   # check inputs
   if (is.null(X_pred) || is.null(Y) || is.null(id.vec)) { stop("You forgot some inputs to fitGLMM().") }
   if (is.unsorted(id.vec)) { stop("Your data must be ordered by subject, please do so before running fitGLMM().") }
@@ -74,10 +76,41 @@ fitGLMM <- function(X_pred = NULL,
                                      })
     marge_style_names <- glm_marge_knots$old_coef
     coef_names <- glm_marge_knots$coef
+    if (verbose) {
+      message(paste0("Generated a total of ",
+                     length(coef_names),
+                     " basis functions across ",
+                     length(unique(id.vec)),
+                     " subjects."))
+    }
+    # run NB LASSO with all possible basis functions
+    lasso_formula <- stats::as.formula(paste0("Y ~ ", paste0(colnames(glmm_basis_df), collapse = " + ")))
+    if (is.null(Y.offset)) {
+      pruned_model <- mpath::glmregNB(lasso_formula,
+                                      data = glmm_basis_df,
+                                      parallel = FALSE,
+                                      nlambda = 50,
+                                      alpha = 1,
+                                      standardize = TRUE,
+                                      trace = FALSE,
+                                      link = log)
+    } else {
+      pruned_model <- mpath::glmregNB(lasso_formula,
+                                      data = glmm_basis_df,
+                                      offset = log(1 / Y.offset),
+                                      parallel = FALSE,
+                                      nlambda = 50,
+                                      alpha = 1,
+                                      standardize = TRUE,
+                                      trace = FALSE,
+                                      link = log)
+    }
+    # identify nonzero basis functions in minimum AIC model
+    nonzero_coefs <- which(as.numeric(pruned_model$beta[, which.min(pruned_model$aic)]) != 0)
     # build formula automatically
     mod_formula <- stats::as.formula(paste0("Y ~ ",
-                                            paste(colnames(glmm_basis_df), collapse = " + "),
-                                            " + (1 + ", paste(colnames(glmm_basis_df), collapse = " + "),
+                                            paste(colnames(glmm_basis_df)[nonzero_coefs], collapse = " + "),
+                                            " + (1 + ", paste(colnames(glmm_basis_df)[nonzero_coefs], collapse = " + "),
                                             " | subject)"))
     glmm_basis_df <- dplyr::mutate(glmm_basis_df,
                                    Y = Y,
diff --git a/man/fitGLMM.Rd b/man/fitGLMM.Rd
index f49653e..a74b714 100644
--- a/man/fitGLMM.Rd
+++ b/man/fitGLMM.Rd
@@ -13,7 +13,8 @@ fitGLMM(
   approx.knot = TRUE,
   M.glm = 3,
   return.basis = FALSE,
-  return.GCV = FALSE
+  return.GCV = FALSE,
+  verbose = FALSE
 )
 }
 \arguments{
@@ -33,7 +34,8 @@ fitGLMM(
 
 \item{return.basis}{(Optional) Whether the basis model matrix (denoted \code{B_final}) should be returned as part of the \code{marge} model object. Defaults to FALSE.}
 
-\item{return.GCV}{(Optional) Whether the final GCV value should be returned as part of the \code{marge} model object. Defaults to FALSE.}
+\item{return.GCV}{(Optional) Whether the final GCV value should be returned as part of the \code{marge} model object. Defaults to FALSE.
+param vebose (Optional) Should intermediate output be printed to the console? Defaults to FALSE.}
 }
 \value{
 An object of class \code{marge} containing the fitted model & other optional quantities of interest (basis function matrix, GCV, etc.).