Estimating Precision; Sensitivity - Hach DR2400 Manual

Chemical Analysis
3.4.3 Precision

3.4.4 Estimating Precision

3.4.5 Sensitivity

Chemical Analysis
Page 44
each standard and use the corrected standard values to calculate the average
and standard deviation used in the MDL.
Every measurement has some degree of uncertainty. Just as a ruler with
markings of 1 mm leaves some doubt as to the exact length of a measurement,
chemical measurements also have some degree of uncertainty. The quality of the
entire calibration curve determines the precision.
Uncertainty in chemical measurements may be due to systematic errors and/or
random errors. A systematic error is a mistake that is always the same for every
measurement made. For example, a blank can add to each measurement for a
specific compound, giving consistently high results (a positive bias). Random
errors are different for every test and can add either a positive or negative
variation in response. Random errors are most often caused by variation in
analytical technique. Hach chemists work hard to eliminate systematic errors in
Hach procedures using Hach reagents, but response variation occurs in all
chemical measurements.
The method performance section in each procedure provides an estimate of the
procedure's precision. Two types of estimates are used throughout thismanual.
Most of the procedures use a replicate analysis estimate, based on real data. For
precision determined in this manner, the 95% confidence interval of the
distribution is reported. Some newer procedures use a 95% or 99% confidence
interval, which is based on the calibration data for that particular chemistry.
In replicate analysis, a Hach chemist prepares a specific concentration of the
analyte in a deionized water matrix. The standard is then analyzed seven
individual times on a single instrument with the two reagent lots originally used
in the calibration (a total of 14 samples). A standard deviation of each of the two
sets of seven values is calculated, and the worst-case 95% confidence interval of
the distribution is reported in the method. The reported value provides an
estimate of the "scatter" of results at a particular point in the calibration curve.
In cases where precision is determined directly from the calibration data, an
estimate is obtained from the calibration data itself, with no additional replicate
analyses. In this case, the precision is the 95 or 99% confidence interval for the
stated concentrations. The precision range is an estimate of the average response
variation and is based on multiple reagent lots and instruments used in the
calibration. Therefore, it will not exactly predict the true precision range for each
reagent lot, but does provide a useful estimate.
In either case, it is important to realize that the estimates are based on a
deionized water matrix. Precision on real samples with varying matrices can be
quite different from these estimates.
If the concentration obtained from running a standard solution does not fall
within the stated precision, please refer to 3.4.3 Precision. Troubleshooting a Test
When Results are in Doubt.
Hach's definition of sensitivity is the change in concentration (∆Concentration)
for a 0.010 change in absorbance (∆Abs).
5940022ChemAnal.fm