Basic HTML Version
22
Bias
Bias is systematic or persistent distortion of a measurement process that causes errors in one direction.
Bias may originate from sources such as calibration errors, response factor shifts, unaccounted-for
interferences, or chronic sample contamination. The sample itself may generate real or apparent bias
caused by a matrix effect or variation in physical properties such as particle size. A bias DQI is a
quantitative indicator of the magnitude of systematic error resulting from these effects. Bias can be in the
positive (high) or negative (low) direction from the true value and is usually unknown in magnitude.
MERC estimates bias by testing the measurement system result against a specimen with known
properties. The most common DQIs for bias are derived from the results of QC samples such as spiked
samples, standard reference materials, and various kinds of blanks in the sample stream. Table 4 lists
some common MERC QC samples and the components of bias they are intended to measure.
Table 4. QC samples for deriving bias indicators.
Sample Type
Indicator For
Blank spike
Instrument contamination or malfunction, calibration shift
Matrix spike
plus
effectiveness of sample extraction/digestion procedures
Reference material
Same as matrix spike, but more representative of overall performance
when material is similar to matrix examined in the study
Calibration blank
Instrument contamination, calibration shift
Preparation blank
plus
laboratory contamination
Field blank (equipment/trip)
plus
field, transportation, and storage contamination
The difference between the measured and expected result is a DQI for bias. For spikes and reference
materials, MERC expresses bias as a fractional or percent comparison of the measured result to the
expected result.
The percent spike recovery is calculated as follows:
percent spike recovery = (x
s
– x
u
) x 100%
x
a
where x
s
= measured value of spiked sample
x
u
= measured value of unspiked sample
x
a
= known amount of spike in sample
A completely unbiased result thus has recovery of 1 (or 100%) and recovery may be greater or less than 1
(100%) depending on whether the result is higher or lower than the known quantity. For blank samples,
the actual magnitude of the result is the DQI because the "known" quantity should be zero for a blank.