如何在R语言中为每个受试者重复执行分析代码并将结果追加至centralities数据框？

Alright, let's figure out how to run this repeated analysis for every subject efficiently in R. I'll walk you through two reliable methods—one using base R, and another with the tidyverse for a more streamlined workflow.

Base R Loop (Straightforward & Compatible) #

First, we'll avoid the inefficient row-by-row appending you were using (it forces R to reallocate memory every time, which slows things down with large datasets). Instead, we'll store results in a list first, then combine them at the end.

1. Grab all unique subject IDs from your dataset:

   unique_subjects <- unique(fullDataDetrend$subjno)
   

2. Initialize an empty list to hold results for each subject:

   centrality_results <- list()
   

3. Loop through each subject, run your analysis, and store the output:

   for (subj in unique_subjects) {
     # Subset data for the current subject
     fullDataDetrend_subj <- subset(fullDataDetrend, subjno == subj, 
                                    select = c(subjno, depressed, sad, tired, interest, happy, neg_thoughts, concentration_probl, ruminating, activity, datevar, timestamp, dayno, beepno))
     
     # Fit the VAR model
     model_subj <- var1(fullDataDetrend_subj)
     
     # Extract the omega_zeta matrix
     model_subj_omega <- getmatrix(model_subj, "omega_zeta")
     
     # Calculate centrality metrics
     centrality_model_subj_omega <- centrality(model_subj_omega)
     
     # Store results in the list (as a data frame for easy combining later)
     centrality_results[[as.character(subj)]] <- data.frame(
       subject_id = subj,
       InExpectedInfluence = centrality_model_subj_omega$InExpectedInfluence
     )
   }
   

4. Combine all list entries into your final centralities data frame:

   centralities <- do.call(rbind, centrality_results)
   

Tidyverse + purrr (Cleaner Pipeline) #

If you prefer a more readable, pipeline-style approach, the tidyverse's dplyr and purrr packages make this a breeze.

1. Load the required packages first:

   library(dplyr)
   library(purrr)
   

2. Use nesting and map functions to run the analysis per subject:

   centralities <- fullDataDetrend %>%
     # Keep only the columns you need
     select(subjno, depressed, sad, tired, interest, happy, neg_thoughts, concentration_probl, ruminating, activity, datevar, timestamp, dayno, beepno) %>%
     # Group data by subject
     group_by(subjno) %>%
     # Nest each subject's data into a list column
     nest() %>%
     # Run each step of your analysis on the nested data
     mutate(
       model = map(data, ~var1(.x)),
       omega_matrix = map(model, ~getmatrix(.x, "omega_zeta")),
       centrality = map(omega_matrix, ~centrality(.x)),
       # Extract the specific metric you need
       InExpectedInfluence = map_dbl(centrality, ~.x$InExpectedInfluence)
     ) %>%
     # Keep only the columns you care about
     select(subjno, InExpectedInfluence) %>%
     rename(subject_id = subjno) %>%
     ungroup()
   

Pro Tips for Robustness #

• Handle errors gracefully: Some subjects might have insufficient data for the var1 model. Add error handling to skip problematic cases and log issues:
  • In base R, use tryCatch() inside the loop:

    for (subj in unique_subjects) {
      tryCatch({
        # Your analysis code here
        centrality_results[[as.character(subj)]] <- data.frame(...)
      }, error = function(e) {
        message(paste("Skipping subject", subj, "due to error:", e$message))
      })
    }
    

  • In tidyverse, use safely() to wrap your analysis function and capture errors:

    safe_centrality <- safely(function(data) {
      model <- var1(data)
      omega <- getmatrix(model, "omega_zeta")
      centrality(omega)$InExpectedInfluence
    })
    
    centralities <- fullDataDetrend %>%
      select(...) %>%
      group_by(subjno) %>%
      nest() %>%
      mutate(
        result = map(data, safe_centrality),
        InExpectedInfluence = map_dbl(result, "result"),
        error_message = map_chr(result, ~ifelse(is.null(.x$error), NA_character_, .x$error$message))
      ) %>%
      select(subjno, InExpectedInfluence, error_message) %>%
      ungroup()
    

内容的提问来源于stack exchange，提问作者Brage Kraft