CodeSiteEff_I2_par_usage • MUGS

Introduction

The CodeSiteEff_I2_par function estimates site-specific effects using input embeddings and penalization methods. This vignette demonstrates how to utilize the function with appropriate input data and parameters.

Load the Required Library

Ensure the MUGS package is loaded before running the example:

library(MUGS)

Load the Data

Load the required datasets for the example:

# Load required data
data(S.1)
data(S.2)
data(X.group.source)
data(X.group.target)
data(U.1)
data(U.2)

Prepare Variables

Set up the variables required for the CodeSiteEff_I2_par function:

# Set parameters
n1 <- 100
n2 <- 100
p <- 5

# Ensure row and column names are consistent for matching
rownames(U.1) <- as.character(seq_len(nrow(U.1)))  # "1" to "100"
rownames(U.2) <- as.character(seq(from = 51, length.out = nrow(U.2)))  # "51" to "150"

# Align S.1 and S.2 with embeddings
rownames(S.1) <- rownames(U.1)
colnames(S.1) <- rownames(U.1)

rownames(S.2) <- rownames(U.2)
colnames(S.2) <- rownames(U.2)

# Get common codes
names.list.1 <- rownames(S.1)
names.list.2 <- rownames(S.2)
common_codes <- intersect(names.list.1, names.list.2)
n.common <- length(common_codes)

if (n.common == 0) stop("Error: No common codes found between source and target sites.")

full.name.list <- c(names.list.1, names.list.2)

# Initialize delta matrix
delta.int <- matrix(0, length(full.name.list), p)
rownames(delta.int) <- full.name.list

Run the Function

Run the CodeSiteEff_I2_par function:

# Estimate site-specific effects
CodeSiteEff_l2_par.out <-  CodeSiteEff_l2_par(
  S.1 = S.1,
  S.2 = S.2,
  n1 = 100,
  n2 = 100,
  U.1 = U.1,
  U.2 = U.2,
  V.1= U.1,
  V.2 = U.2,
  delta.int = delta.int,
  lambda.delta = 3000,
  p=5,
  common_codes = common_codes,
  n.common = 50,
  n.core=2)
#> Warning: package 'doSNOW' was built under R version 4.4.1
#> Loading required package: foreach
#> Warning: package 'foreach' was built under R version 4.4.1
#> Loading required package: iterators
#> Loading required package: snow
#> 
#> Attaching package: 'snow'
#> The following objects are masked from 'package:parallel':
#> 
#>     closeNode, clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
#>     clusterExport, clusterMap, clusterSplit, makeCluster, parApply,
#>     parCapply, parLapply, parRapply, parSapply, recvData, recvOneData,
#>     sendData, splitIndices, stopCluster

Examine the Output

Explore the structure and key components of the output:

# View structure of the output
str(CodeSiteEff_l2_par.out)
#> List of 3
#>  $ delta  : num [1:200, 1:5] 0 0 0 0 0 0 0 0 0 0 ...
#>   ..- attr(*, "dimnames")=List of 2
#>   .. ..$ : chr [1:200] "1" "2" "3" "4" ...
#>   .. ..$ : NULL
#>  $ V.1.new: num [1:100, 1:5] 0.206 1.437 0.28 0.71 -0.543 ...
#>   ..- attr(*, "dimnames")=List of 2
#>   .. ..$ : chr [1:100] "1" "2" "3" "4" ...
#>   .. ..$ : NULL
#>  $ V.2.new: num [1:100, 1:5] 0.468 1.595 -0.152 -1.13 -0.165 ...
#>   ..- attr(*, "dimnames")=List of 2
#>   .. ..$ : chr [1:100] "51" "52" "53" "54" ...
#>   .. ..$ : NULL

# Print specific components of the result
cat("\nEstimated Effects (Delta):\n")
#> 
#> Estimated Effects (Delta):
print(CodeSiteEff_l2_par.out$delta[1:5, 1:5])  # First 5 rows and columns of delta matrix
#>   [,1] [,2] [,3] [,4] [,5]
#> 1    0    0    0    0    0
#> 2    0    0    0    0    0
#> 3    0    0    0    0    0
#> 4    0    0    0    0    0
#> 5    0    0    0    0    0

cat("\nRegularization Path:\n")
#> 
#> Regularization Path:
print(CodeSiteEff_l2_par.out$path)
#> NULL

Notes

Custom Parameters: Modify parameters like n1, n2, p, and lambda.delta to test different scenarios.
Data Preparation: Ensure datasets (S.1, S.2, U.1, U.2, etc.) are correctly loaded and aligned.
Output: Key components include the estimated delta matrix and the regularization path.

Summary

This vignette demonstrated how to use the CodeSiteEff_l2_par function for estimating site-specific effects. Adjust input parameters and datasets to test different scenarios and interpret the output components for your analysis.