Changes in version 0.0.0.9000                      

Checks

  - Every check now carries a plain-language rule and guidance on what
    to do when it flags; check_catalogue() lists them and the app shows
    them next to each finding.
  - run_nmpk_checks() returns a title column, and partial thresholds
    lists fall back to default_thresholds(). New thresholds:
    outlier_iqr_k and predose_cmax_frac.
  - New findings: DV or MDV = 0 on dose records (CORE-OBS-001), DV below
    LLOQ without a BLQ flag (CORE-OBS-002), non-numeric ID
    (CORE-STRUCT-002), missing EVID (CORE-STRUCT-003), ID versus USUBJID
    uniqueness (CORE-INT-001), exclusion reason on a record that is not
    excluded (CORE-MR-001), ADDL on non-dose records (CORE-DOSE-005),
    and placeholder RATE values or AMT / RATE / DUR disagreement
    (CORE-DOSE-006).
  - Fewer false positives: outliers are compared within nominal time
    point and analyte (CORE-OBS-005, CORE-TIME-004), BLQ troughs before
    the next dose are not "mid-profile" (CORE-OBS-003), ordinary dose
    levels are not outliers (CORE-DOSE-002), TIME may restart on reset
    records (CORE-TIME-001), and time-varying covariates are no longer
    reported as inconsistent fixed covariates (CORE-COV-001,
    CORE-COV-006).
  - Changed rules: predose concentrations are compared with 5% of the
    subject's Cmax (CORE-OBS-004); covariate outliers use boxplot fences
    and cover more covariates (CORE-COV-004); missingness is reported by
    event type and flags only unexpected gaps (CORE-STRUCT-006);
    duplicates are split into exact rows and conflicting values
    (CORE-STRUCT-005); cross-study medians are compared from two studies
    upwards (CORE-INT-003); CORE-MR-003 lists the subjects flagged by
    other checks.
  - USUBJID and VISIT are canonical variables; USUBJID is no longer
    preferred over a numeric identifier for ID.

App

  - Built-in example datasets can be loaded from the Data tab.
  - Uploads accept comma-, semicolon-, tab-, and space-delimited text
    and read ., the NONMEM null value, as missing.
  - Findings is the main workspace: it lists the checks to review with
    triage progress, and for the selected check shows the rule,
    guidance, the profile of each affected subject, all result tables,
    and the triage decision, which is saved as it is entered.
  - Overview describes the dataset before any check runs: all
    concentration-time profiles (by time since first dose or after the
    last dose), dose levels, and missing values by event type.
  - The sidebar is replaced by a status line with the dataset, the state
    of the last run, and a run button; thresholds moved to the Data tab.
  - Profiles can be filtered by check; new Check library tab.
  - Results are cleared when another dataset is confirmed; a column can
    no longer be mapped to two variables.
  - New compact, neutral theme. The Data tab shows the three steps next
    to the data preview; the status line stays visible while scrolling;
    the Findings tab shows the number of checks to review; severity,
    result, and triage state are shown as badges; plots use a
    colour-blind safe palette.