Data Quality and Quality Assurance
What this domain tests
This domain tests the systematic planning and verification of environmental data. Expect items on data quality objectives, quality assurance project plans, quality control samples, data validation, and the interpretation of precision, accuracy, representativeness, comparability, completeness, and sensitivity.
Core rules and concepts
- Data quality objectives define the decision, the data needed, the tolerable error, and the sampling design before data are collected.
- Quality assurance project plans document project management, measurement methods, quality control, and corrective action.
- Quality indicators. Precision is agreement among replicates, often expressed as relative percent difference. Accuracy or bias is agreement with a reference or spike recovery. Representativeness is how well data reflect the population. Comparability allows data sets to be combined. Completeness is the percentage of valid data obtained against the plan. Sensitivity is the ability to detect at the required level.
- Detection limit terms differ. The method detection limit is a statistical detection capability; the reporting limit is the lowest calibrated quantified value; neither should be treated as zero when nondetect.
- Validation checks holding times, preservation, blanks, duplicates, spikes, calibration, and transcription before data are used.
Worked example
Duplicate results are 12.4 and 14.8 micrograms per liter. The relative percent difference is the absolute difference divided by the mean, times 100: 2.4 divided by 13.6, about 18 percent. If the project criterion is 20 percent at that concentration, precision passes. Now suppose the method blank contains 2.1 micrograms per liter. The blank is below the sample result but may indicate low-level contamination; the validator must decide whether the blank exceeds the acceptance rule and whether affected results need qualification. The correct answer identifies both checks and does not simply average the duplicates.
Traps that cost marks
Reporting nondetects as zero; confusing the method detection limit with the reporting limit; evaluating precision without the project criterion; ignoring blank contamination because samples are higher; and calculating completeness without excluding rejected data.