Details for function DataCheck
In PDRobust: Robust Longitudinal Effects Under Truncation by Death

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  eval = TRUE,
  echo = TRUE,
  include = TRUE,
  fig.show = "asis",
  fig.keep = "high",
  warning = FALSE,
  message = FALSE
)

```{=html} ```

Columns in the itemized checks

The component check$checks is a data frame with the following columns:

| Column | Description | |:---|:---| | check | The unique name of the validation check. | | passed | Indicates whether the dataset passed the corresponding check. | | severity | Classifies the result as "error", "warning", or "information". | | standardize_can_fix | Indicates whether the issue can be handled automatically by DataStandard(), potentially with drop = TRUE. | | requires_manual_resolution | Indicates whether the issue must be corrected or reviewed manually before standardization or analysis. | | analysis_blocking | Indicates whether failure of the check prevents the dataset from being considered analysis-ready. | | details | Provides a concise summary of the observed result, including relevant counts, values, rows, subjects, or time points. | | recommendation | Describes the recommended action for resolving or interpreting the result. |

::: {.validation-check-table}

| Check | What is evaluated | Corresponding report and diagnostics | |----|----|----| | required_columns | Verifies that all structural variables and mapped covariates are present in the input dataset. | Reports the number of required columns or lists the missing columns. Missing columns require manual correction of the data or mapping. | | nonempty_data | Verifies that the input dataset contains at least one row. | Reports the number of detected rows. An empty dataset requires manual resolution. | | missing_id_or_time | Identifies records with missing subject identifiers or time values. | Reports the affected original row numbers and stores them in diagnostics$missing_id_time_rows. These rows may be removed using drop = TRUE when appropriate. | | time_encoding | Determines whether the time variable is numeric or can be converted unambiguously to numeric values. | Reports the original class and any invalid rows. Unambiguous character or factor encodings can be converted during standardization; ambiguous values require manual correction. | | mapping_time_endpoints | Verifies that baseline_time and cutoff_time are observed in the data and that baseline_time <= cutoff_time. | Reports the mapped endpoints and all observed time values. Invalid endpoints require correction of the mapping or time coding. | | analysis_time_grid | Identifies all observed time points within the mapped baseline-to-cutoff window and verifies that both endpoints are included. | Reports the complete retained analysis-time grid. Every observed visit within the mapped window is included in the standardized grid. | | duplicate_id_time_records | Detects duplicated subject–time combinations. | Reports the number of duplicated rows and affected subjects. Detailed values are stored in diagnostics$duplicate_rows and diagnostics$duplicate_subjects. Duplicates require a manually specified aggregation or record-selection rule. | | complete_longitudinal_structure | Determines whether each subject has a usable record at every retained analysis time. Duplicate records are evaluated separately. | Reports the number and percentage of complete subjects, the number of incomplete subjects, and missing-subject counts by time. Details are stored in diagnostics$incomplete_subjects and diagnostics$missing_by_time. | | treatment_encoding | Verifies that treatment is completely observed and encoded as binary 0/1, or through an explicitly convertible binary representation. | Reports the variable class, observed values, missing rows, invalid rows, invalid values, and affected subjects. Detailed results are stored in diagnostics$treatment_invalid_rows and diagnostics$treatment_invalid_subjects. | | survival_encoding | Verifies that survival status is completely observed and encoded as binary 0/1, or through an explicitly convertible binary representation. | Reports the variable class, observed values, missing rows, invalid rows, and invalid values. Invalid-row indices are stored in diagnostics$survival_invalid_rows. | | treatment_consistency_within_subject | Verifies that baseline treatment assignment remains constant within each subject over follow-up. | Reports the number and identifiers of subjects whose treatment value changes. Affected identifiers are stored in diagnostics$treatment_changes. | | survival_consistency_within_subject | Verifies that a subject does not transition from S = 0 back to S = 1 at a later time. | Reports the number and identifiers of subjects with impossible survival transitions. Affected identifiers are stored in diagnostics$impossible_survival_transitions. | | outcome_type_and_encoding | Verifies that the outcome agrees with mapping$y_type: binary outcomes must use valid binary values, whereas continuous outcomes must contain finite numeric values. | Reports the original class, observed values, and invalid or non-finite rows. Invalid-row indices are stored in diagnostics$outcome_invalid_rows. Unambiguous conversions are performed by DataStandard(). | | structural_outcome_missingness_after_death | Counts records for which S = 0 and Y = NA. | Returns an informational result reporting the number and percentage of structurally missing outcomes. These outcomes are expected and should not be imputed or replaced with observed zeros. | | outcome_observed_after_death | Identifies records for which S = 0 but the outcome remains observed. | Reports the number and original row numbers of such records. Row indices are stored in diagnostics$outcome_observed_after_death_rows. Manual verification is required. | | outcome_missingness_among_survivors | Identifies ordinary outcome missingness among records with S = 1. | Reports the number of missing records, affected subjects, and counts by analysis time. Details are stored in diagnostics$outcome_missing_alive_rows and diagnostics$outcome_missing_alive_by_time. | | missing_covariates | Evaluates missingness in every mapped covariate. | Reports, for each covariate, the number of missing records, number of affected subjects, and percentage of affected subjects. The complete summary is stored in diagnostics$covariate_missing. | | time_coding_and_order | Determines whether rows are ordered by subject and time and whether the analysis-time grid already uses consecutive integers beginning at zero. | Reports the observed raw time values and whether records are correctly ordered. DataStandard() can sort the records and map the retained time grid to 0, 1, ..., n. | | id_coding | Determines whether subject identifiers are consecutive integers and whether records are correctly ordered. | Reports the identifier class, number of unique subjects, and whether the coding is canonical. DataStandard() creates an ID audit map and assigns consecutive integer identifiers. | | treatment_group_availability | Verifies that both treatment groups are represented at baseline. | Reports the number of unique baseline subjects in treatment groups 0 and 1. Group counts are stored in diagnostics$treatment_group_counts. | | covariate_variation | Identifies mapped covariates that are constant or have near-zero variation. | Reports the names of problematic covariates and stores them in diagnostics$near_zero_variation_covariates. This is a nonblocking warning, but the covariates should be reviewed before model fitting. | | retained_sample_after_optional_dropping | Records the sample size before optional subject-level deletion during standardization. | Initially reports the number of subjects present. After DataStandard(), the corresponding report is updated with the original, removed, and retained subject counts and the retained percentage. |