In geochemical analysis, bias is a consistent directional deviation of measured values from the true reference value.
Unlike random error, it cannot be reduced by increasing replicates. It must be identified, characterized, and corrected at its source.
Most common sources of systematic bias in exploration and mining datasets:
1. Incomplete Digestion
Partial dissolution of refractory minerals such as chromite, zircon, and ilmenite results in a systematic low bias for elements hosted in those phases. For example, multi-acid digestions without HF will consistently under-report elements in silicate matrices.
2. Matrix-Matched Calibration Failure
Calibrating high-TDS samples against aqueous standards leads to signal suppression or enhancement. This matrix effect introduces a persistent bias across the entire batch.
3. Spectral Interference
Unresolved peaks in ICP-OES or isobaric and polyatomic interferences in ICP-MS create a systematic high bias. A classic example is BaO+ interference on Eu in ICP-MS.
4. Instrumental Drift
Gradual signal change during extended analytical runs due to warm-up instability, nebulizer blockage, or plasma energy variation.
How to Detect It:
When CRM results consistently plot above or below the certified value across multiple batches, and the deviation persistently exceeds ±2 sigma, you are looking at systematic bias.
A 97% recovery on every batch is not “acceptable.” It is a persistent 3% negative bias that will impact your resource model.
How to Correct It:
Best practice is prevention through appropriate multi-acid digestion protocols and matrix-matched standards.
Bias correction factors should only be applied when the bias is quantified, stable, and traceable to an identified root cause. Applying a blanket correction factor without understanding the cause is analytically inappropriate.
Reliable data drives reliable decisions.
If you are seeing persistent bias in your QAQC data, the expert team at Orelysis can help you diagnose the root cause and implement corrective actions.
Leave a Reply