Main steps in data analysis

A simple workflow

Key functionalities include wrnet() for pathwise solution of $\beta$ as a function of $\lambda$, cv_wrnet() for $k$-fold cross-validation, and test_wrnet() for evaluation of test performance. Additionally, there are helper functions to calculate or visualize C-indices, variable importance, and more.

Data preparation

Initial data frame df in long format with id, time, status columns: subject identifiers, event times, and event status (1 for death; 2 for nonfatal event; 0 for censoring), along with covariate columns.
Partition into training and test sets by a specified proportion:
```
      wr_split(df, prop = 0.8)
```
- Returns a list of training data df_train and test data df_test;
- Default split stratified by outcome status.

Cross-validation

Perform $k$-fold cross-validation on df_train:
```
       cv_wrnet(id, time, status, Z, k = 10, ...) 
```
- Returns a tibble with columns lambda and concordance;
- Identify optimal lambda_opt maximizing concordance;
- Additional arguments, such as $\alpha\in[0, 1]$ (default is 1), passed to glmnet::glmnet() for pairwise logistic regression on each fold.

Fit final model

Final fit on df_train under optimal lambda_opt:
```
    final_fit <- wrnet(id, time, status, Z, lambda = lambda_opt, ...)
```
- $\widehat\beta(\lambda_{\rm opt})$: final_fit$beta;
- Variable importance: vi_wrnet(final_fit);
- Additional arguments passed to glmnet::glmnet().

Test performance

Calculate overall and component-wise C-indices on df_test:
```
      test_wrnet(final_fit, df_test)
```
- Returns an object with element concordance, a tibble containing the calculated metrics.