1	Uncertainties in Dilution Practices and Calibrations©2006
2	Protocol/SPEX We manufacture organic and inorganic certified reference materials (CRM’s) We also manufacture sample preparation equipment. We have an ISO 9001-2000 quality system We are NVLAP approved to manufacture Proficiency Testing Samples. We are Audited and accepted as contractors to various state and federal agencies throughout the world. We were the first in the world to make, market and sell elemental calibration standards for OES……
3	Why talk about Uncertainties in Calibrations and Dilutions at all? We have observed over the years that chemists use a variety of techniques to dilute solutions and calibrate instruments. We have also observed that many techniques in common practice contain an unacceptable amount of error and uncertainty. We wish to promote good laboratory practices and reduce the errors at the bench. The goal is to show calculations, uncertainties and lab practices that will help chemists make better measurements, or understand why they do so well.
4	Why are so many people in the lab business suddenly so concerned about uncertainties? Evaluators of data have always been concerned with the uncertainty of a result Requirements for analyzing uncertainties emerged with ISO 17025 A point was put on it for us when we entered the NIST/NVLAP Providers of PT program
5	What is uncertainty? Uncertainty is a parameter, associated with the result of a measurement that defines the range of the values that could reasonably be attributed to the measured quantity. For the purposes of this talk, the expected variance around a calibration or test When uncertainty is evaluated and reported in a specified way it indicates the level of confidence that the value actually lies within the range defined by the uncertainty interval
6	What contributes to uncertainty Any measurement is subject to imperfections Random effects, such as short-term fluctuations in temperature, humidity and air-pressure Variability in the performance of the measurer Uncertainty of the value of a reference standard Sampling…..
7	Other contributors to uncertainty Other imperfections are due to the practical limits to which correction can be made for systematic effects, such as offset of a measuring instrument, drift in its characteristics between calibrations or personal bias in reading an analog scale
8	Techniques This presentation uses examples most commonly used in GC, GC/MS, ICP and ICP/MS techniques The same calibration principles apply to most analytical techniques The dilution schemes are specific to GC practices Repeated measurements will show the variation due to random effects
9	Calibration Strategies There are many ways to Calibrate an instrument. 1 point 2 point Multiple points External Standard Internal Standard
10	Do more points = more confidence? Generally yes, but there is a point of diminishing returns Calibrating below or above your working range skews your curves Too many points, over time, are meaningless and expensive
11	External vs Internal Standard External standard techniques are highly dependent on consistency in sample prep, sample introduction, dilutions and instrument conditions Internal standard techniques give you a relative response to quantitate and is generally much more accurate. Effective only within the same order of magnitude.
12	Syringes and Pipettes Always use the same syringe or pipettor for I/S additions Check your accuracy and precision Check the syringes, pipettes and volumetrics for accuracy and precision Make sure that all analysts are being consistent
13	Internal Standard Variations
14	Typical linear calculations Average response factor Linear Regression Least squares
15	Average Response Factors Sum of all response factors / n Is a weighted way to evaluate a straight line Used in CLP, 8000, 500 and 600 series methods CLP Organics Daily Cal range = 25 % rsd for 50 STD This means that from 37.5 to 62.5 passes Is their a national accreditation body in the UK? NELAC, in the states, is now requiring an estimate of Uncertainty
16	Linear Regressions y=mx+b m=slope b=intercept r=Correlation Coefficient r²=The square of r is conventionally used as a measure of the strength of the association between X and Y. For example, if the coefficient is .90, then 81% of the variance of Y is said to be explained by the changes in X and the linear relation between X and Y within the range observed.
17	Linear Regression Correlation Coefficient
18	Typical Naphthalene Curve
19	What does it look like?
20	A closer look…
21	Just the facts…
22	Uncertainty in dilutions Without exploring the uncertainties of our measurements, we cannot have confidence in our calibrations and therefore our results. Most labs have NEVER analyzed their uncertainties. Analyzing your uncertainties allows you to analyze the weaknesses in your systems. In this case uncertainties show you more accurate ways to dilute.
23	Dilution Schemes There are many ways to perform dilutions Using uncertainties, you can determine the best methods Always use dilutions appropriate to the scale of your application There is always the good, the bad and the ugly…. In reverse, they are
24	The Ugly… How many people create matter in sludge/solids? Remember those 1,200,000mg/kg results for Calcium? Huge dilutions = HUGE expanded error
25	The Bad Using a 10 ul syringe to dilute 2 ul to 1 L is bad The Expanded error at 2 ul = 23% A target of 2 ug/L from a 1000 ug/ml std using 2 ul = 1.77 to 2.23ug/L.
26	The Good Use dilutions that are within the 100x level of your working volumes, when using syringes and 1000x when using pipettes Use at least 50% of your syringe volumes Use calibrated pipettes and glassware
27	An experiment in syringe use and uncertainty We had 7 people weigh various volumes with various size syringes. There were 7 observations at each volume.
28	Measure of liquid from syringes (1 of seven)
29	Mean syringe observations (H₂O) and Uncertainties
30	Mean syringe observations and Uncertainties (Methanol)
31	Mean syringe observations and Uncertainties (Methylene Chloride)
32	Mean pipettor and pipette observations and Uncertainties (H₂O)
33	A Frequently used point dilution scheme
34	Cumbersome…… Without total isolation while performing these dilutions, errors are almost guaranteed Add another 1.11% if the I/S was inclusive of the 1 ml volume Combining all of the uncertainties illustrates a 3.03% curve uncertainty
35	Serial Dilutions
36	Serial Dilutions (continued) Combining all of the uncertainties illustrates a 1.28% curve variability Add another 1.11% if the I/S was inclusive of the 1 ml volume (Total =2.39%) Offer more accuracy (21% less uncertainty) This will cause less failures (~5% in PT’s) and more accurate results
37	Serial Dilutions (continued) Dilutions should not exceed 1000x at the 1Liter volume (error = 1.16%). The error is generally acceptable here. If you dilute in excess of 10,000x, then your analytical uncertainties become astronomical!
38	Uncertainties in Methods Aside from sampling uncertainty and matrix issues, the largest contributors to method uncertainties are the combination of the daily calibration deviation and analyte recovery data. Examples are: CLP Daily Cal = 25 % rsd This means that from 37.5 to 62.5 passes Therefore, a result of 110 ug/L may be anywhere between 82.5 and 137.5!
39	Let’s take this to 8270C…. You can develop recovery uncertainties from your LCS data Let’s consider Phenol recovery at 5%-112% 1,4-Dichlorobenzene recovery at 20%-124% Daily Cal at 25 % Phenol results for a 100 ug/L PT are anywhere from 3.75 to 137!!!! 1,4-Dichlorobenzene results for a 100 ug/L PT are anywhere from 15 to 149!!!!
40	8270 (continued) This is real data published in 8270C With NELAC’s new standards calling for uncertainties, expect legal issues to arise For an end user of data, seeing a 9ppb result for 1,4-Dichlorobenzene and knowing the uncertainties, he has to wonder if the real concentration is 7 or 60ppb!!
41	Whatever your methods Validate Validate Validate Validate
42	References References ISO Guide 17025: Certification of Reference Materials, general and statistical principles ASTM Guide D6362-98 ILAC-G21-2000 ISO/REMCO N280 EURACHEM/CITAC Guide, Second Edition ISO Guidelines for Uncertainty Calculations for Chemical Analysis: NCSL, Robert Watters and Mark Levenson-NIST Method Validation and Measurement Uncertainty: Advanced Concepts in Analytical Quality Assurance: Wolfhard Wegscheider-Symposium at NIST NIST Technical Note 1297 Guide to the Expression of Uncertainty in Measurement (1995)
43	Thank You! We Want You to Be Successful Call us for technical assistance! 0208 204 6656 (Reference Materials) Email to bhahn@prostds.com or pbutler@spexcertiprep.co.uk