AgentSkillsCN

data-qc

数据集质量控制——针对适配 NONMEM 的数据集进行全面检查

SKILL.md
--- frontmatter
name: data-qc
description: Dataset Quality Control - Comprehensive checks for NONMEM-ready datasets

Dataset QC

Usage

code
/data-qc
/data-qc path/to/dataset.csv

Checks

1. Structure

  • • Required columns present (ID, TIME, DV, AMT, MDV, EVID, CMT)
  • • No unexpected NA values in key columns
  • • Correct data types (numeric for DV, integer for ID)

2. Dosing Records

  • • All subjects have at least one dosing record (EVID=1)
  • • AMT > 0 when EVID = 1
  • • AMT = 0 or missing when EVID = 0
  • • CMT appropriate for dose and observation

3. Observation Records

  • • DV present when EVID = 0
  • • No negative DV values (unless log-transformed)
  • • BLQ handling documented

4. Time

  • • No negative TIME values (except pre-dose if applicable)
  • • TIME sorted within each subject
  • • No duplicate TIME within subject for same EVID

5. Covariates

  • • Values within physiological ranges
    • •WT: 30-200 kg
    • •AGE: 18-100 years
    • •CRCL: 10-200 mL/min
  • • No missing values for key covariates
  • • Categorical covariates properly coded (0/1 or integers)

6. Summary Statistics

Generate and present:

  • •N subjects, N observations, N doses
  • •DV summary (min, max, median, N BLQ)
  • •Covariate summaries
  • •Observations per subject (min, max, median)

Output

Present QC report with PASS/FAIL for each check and summary statistics.