1

SAMPLING AND MEASUREMENT UNCERTAINTY

Sri Noegrohati, UGM

To make an objective, wise, fair, and impartial decision, a valid information of the intended population is required The information could be qualitative or quantitative, therefore scientific data generated through valid measurements is required. Sampling is the first part of the whole process of measurement. Without representative sample with good protection of its integritytherefore sampling is an important part in making decision

ri Noegrhay

Measurements and decisions

3

Wise decision are based on scientific information obtained by valid measurements. Examples : Acceptance of consignments Testing for batch releases Control of raw materials Control of in process products Finished product controls Release of non-conforming products Legal disputes Inter laboratory trials The credibility of these decisions depends on the uncertainty of the measurement results the most important parameter that describes the quality of measurements is uncertainty of measurement Sri Noegrhay Sri Noegrohati, UGM

Sampling and Measurements

4

It is impossible to analyze the entire bulk of the material to be characterized a measurement always involves the process of taking a representative sample the uncertainty associated with the sampling process will contribute to the uncertainty associated with the reported result The uncertainty arising from the sampling process must therefore be evaluated. Since analytical and sampling processes contribute to the uncertainty in the result, the uncertainty can only be estimated if there is an understanding of the complete process.

Sampling planners and analytical scientists optimize the whole measurement procedure, and plan a strategy to estimate the uncertainty reliable decisions based upon the measurements for the customer .
Sri Noegrhay

Sampling and uncertainty of Measurements

5

for making reliable interpretation of measurements, and judging their fitness for purpose, it is important to know the total uncertainty in a measurement To estimate the uncertainty of measurement, arising from the processes of sampling and the physical preparation of samples, it takes a holistic view of the measurement process, including all of the sampling steps as well as the analytical process, describing the effects and errors that cause uncertainty in the final measurement.

Sampling is not just delivering the sample to the laboratory.

Sri Noegrhay

Sampling target
6

sampling : a procedure whereby a part of a substance, material or product is taken to provide for testing or calibration a representative sample of the whole. (ISO/IEC
10725)

Sampling target is Portion of material (the whole of a batch, lot or consignment), at a particular time, that the sample is intended to represent to be characterized Notes:
1. The sampling target should be defined prior to designing the sampling plan. 2. The sampling target may be defined by Regulations: composition of a whole batch sampling target: whole batch 3. If the interest are the properties and characteristics of the certain area or period, then each location will be a separate sampling target.
Sri Noegrhay Sri Noegrohati, UGM

Uncertainty of Primary sample and Composite sample

7

The whole process of measurement begins with the taking of the primary sample from a sampling target. Primary sample: The collection of one or more increments or units initially taken from a population. Note: The term primary, in this case, does not refer to the quality of the sample, rather the fact that the sample was taken during the earliest stage of measurement. primary samples are often combined to form a composite sample before a measurement is made the uncertainty of this single composite sample The value of this uncertainty will be affected by the number of primary samples taken. The resulting sample goes through intermediary steps, such as transportation and preservation of samples, prior to the analytical determination. Each steps contribute to the uncertainty of measurement in the final result, Noegrhay Sri

measurement process diagram

Eurachem/EURO LAB/CITAC/Nordt est Guide, April 2006

Sri Noegrhay

Sri Noegrohati, UGM

Sampling in the measurement process

9

the goal of sampling is to select and obtain a test portion of the material in some manner, such that the sub-sample is representative of the entire experimental unit.

The sampling target that we are studying is not homogenous and the properties vary significant cause of uncertainty both sampling and analysis is associated with uncertainty
Sampling contribute to the uncertainty of measurement describes the quality of measurements
Sri Noegrhay Sri Noegrohati, UGM

The certainty of a decision

10

Any measurement is associated with an uncertainty from : heterogeneity, sampling and analysis. These measurement uncertainty interacts closely with decision making:

define the certainty of a decision shall be made, and control the measurement uncertainty control the certainty of decision through quantitative estimation of the measurement uncertainty including all steps in the chain.
Sri Noegrhay Sri Noegrohati, UGM

11

Sri Noegrohati,UGM

General
12

Sampling should be performed by persons trained in the techniques of sample collection Each lot that is to be examined must be clearly defined. The appropriate Codex Commodity Committee should stipulate how a consignment should be handled in instances where no lot designation exists A lot is a definite quantity of some commodity manufactured or produced under conditions, which are presumed uniform.

For the goods presumed heterogeneous, sampling can only be achieved on each homogeneous part of this heterogeneous lot. In that case, the final sample is called a stratified sample

A consignment is a quantity of some commodity delivered at one time. It may consist in either a portion of a lot, either a set of several lots.
Sri Noegrohati,UGM

Representative sampling
13

The representative sampling is a procedure used for drawing or forming a representative sample A representative sample is a sample in which the characteristics of the lot from which it is drawn are maintained. It is in particular the case of a simple random sample where each of the items or increments of the lot has been given the same probability of entering the sample.

Sri Noegrohati,UGM

Random sampling
14

Random sampling involves the collection of n items from a lot of N items in such a way that all possible combinations of n items have the same probability of being collected. The randomness can be obtained by use of table of random number which can be generated by using computer software. In order to avoid any dispute over the representativeness of the sample, a random sampling procedure should be chosen,
Sri Noegrohati,UGM

Part of Random sampling numbers (ISO 2859-0: table 3)

15

Sri Noegrohati,UGM

stratified random sampling

16

If the lot is heterogeneous, a random sample may not be representative of the lot stratified sampling may be a solution. Stratified sampling consists of dividing the lot into different strata or zones, each stratum being more homogenous than the original lot. Then a random sample is drawn from each of these strata, following specified instructions which may be drafted by the Codex product committees. Each stratum can then be inspected by random sampling which usually includes from 2 to 20 items or increments per sample. it is necessary, where appropriate, to refer to the specific instructions of the Codex product committees.
Sri Noegrohati,UGM

17

When it is not possible to sample at random, it is mandatory:

1. To avoid preferentially choosing items which are more easily accessible or which can be differentiated by a visible characteristic. 2. In the case of periodic phenomena, to avoid sampling every k seconds or every kth package, or every kth centimetres, to take an unit from every nth palette, prepackage,

Sri Noegrohati,UGM

Reliable analytical information

18

Reliable results can only be obtained from samples taken according to the objectives of the study The samples must be representative enable to apply the analytical result to the entire experimental unit. Utmost attention should be given to the selection of sampling methods, handling (packing, labelling, shipping and storage) of samples. Valid analytical results can only be obtained if the samples have been properly taken, dispatched and stored before analysis. The study should be designed to assure the integrity of the whole chain of activities.

SamplingSri Noegrhay plans

Sri Noegrohati, UGM

sampling plan
19

IUPAC (1990), ISO 11074-2 (1998), AMC (2005) Predetermined procedure for the selection, withdrawal, preservation, transportation and preparation of the portions to be removed from a population as a sample. ISO 2859-1(1999) combination of sample size(s) to be used and associated lot acceptability criteria NOTE 1 A single sampling plan is a combination of sample size and acceptance and rejection numbers. A double sampling plan is a combination of two sample sizes and acceptance and rejection numbers for the first sample and for the combined sample. 2 A sampling plan does not contain the rules on how to draw the sample.
Sri Noegrhay

20

Sri Noegrohati,UGM

Primary Samples
21

A primary sample is the portion of product collected from a lot during the first stage of the sampling process, and will normally be in the form of: an item (if collected from a lot of prepacked products) or of an increment (if collected from a bulk lot). However, an increment may be considered to be an item if measurements are made on individual increments. Note: Consider the Nature of the lot Bulk or pre-packed commodities Size, homogeneity and distribution concerning the characteristic to control Nature of the characteristic to control: Qualitative or Quantitative Sri Noegrohati,UGM Nature of the control: Characteristic of individual item or the

precautions
22

As far as is practicable, primary samples should be taken throughout the lot departures from this requirement should be recorded. Sufficient primary samples of similar size should be collected to facilitate laboratory analysis
maintain sample integrity sampling error ssampling i.e., avoid contamination or any other changes

adversely affect the amount of residues or the analytical determinations, or make the laboratory sample not representative of the composite sample from the lot.
Sri Noegrohati,UGM

composite sample
23

When required by the sampling plan, a composite sample is produced by carefully mixing the primary samples items from a lot of pre-packaged products; or increments from a bulk (not pre-packaged) lot. Except for economical reasons, this sampling technique is not to be recommended given the loss of information on sample-to-sample variation due to the combination of primary samples.

Sri Noegrohati,UGM

Final Sample
24

The bulk or bulked sample should, if possible, constitute the final sample and be submitted to the laboratory for analysis. If the bulk/bulked sample is too large, the final sample may be prepared from it by a suitable method of reduction. In this process, however, individual items must not be cut or divided.
National legislative needs may require that the final sample be subdivided into two or more portions for separate analysis. Each portion must be representative of the final sample.
Sri Noegrohati,UGM

Packaging and Transmission of Laboratory Samples

25

laboratory sample : The sample submitted to the laboratory the form of either the final sample or a representative portion of the final sample. should be kept in such a manner that the controlled characteristic is not modified (e.g., for microbiological controls, mandatory use of a sterile and cooled container). should be placed in a clean inert container offering adequate protection from external contamination and protection against damage to the sample in transit.

The container should be sealed in such a manner that unauthorised opening is detectable, sent to the laboratory as soon as possible taking any necessary precautions against leakage or spoilage, e.g., frozen foods should Sri Noegrohati,UGM be kept frozen and perishable samples should be kept cooled or frozen, as appropriate.

Sampling reports
26

Every sampling act implies the drafting of a sampling report and indicating in particular the reason for sampling, the origin of the sample, the sampling method and the date and place of sampling, together with any additional information likely to be of assistance to the analyst, such as transport time and conditions. The samples, in particular the ones for the laboratory, shall be clearly identified. In case of any departure from the recommended sampling procedure necessary to append another detailed report on the deviating procedure which has been actually followed However , this decision is to be taken by the responsible authorities. Sri Noegrohati,UGM

ESTIMATION ERRORS
27

The total standard deviation is given by the formula: where s is the sampling standard-deviation, m the measurement standard-deviation

- the most frequent one: the analytical error << the sampling error, i.e the analytical error is at most equal to one third of the sampling error In this case, m s/3

The standard deviation for the observed results will be at most 5 % larger than the sampling standard deviation taking into account the analytical error. Sri Noegrohati,UGM

Uncertainty
28

The uncertainty arises from a variety of sources, and these can be categorized in different ways Uncertainty is related to other concepts, such as accuracy, error, trueness, bias and precision. Uncertainty is a range of values attributable on the basis of the measurement result and other known effects, whereas error is a single difference between a result and a true (or reference) value Uncertainty includes allowances for all effects that may influence a result (i.e. both random and systematic errors); precision only includes the effects that vary during the observations (i.e. only some random errors).

Sri Noegrhay

Sources of uncertainty
29

Both sampling and analysis is associated with uncertainty Sampling

Heterogeneity (or inhomogeneity)
Effects of specific sampling strategy (e.g. random, stratified random, proportional etc.) Effects of movement of bulk medium (particularly density selection) Physical state of bulk (solid, liquid, gas) Temperature and pressure effects Effect of sampling process on composition (e.g. differential adsorption in sampling system). Contamination Transportation and preservation of sample

Sample preparation
Homogenisation and/or subsampling effects Drying Milling Dissolution Extraction Contamination Derivatisation (chemical effects) Dilution errors (Pre-)Concentration Control of speciation effects
Sri Noegrohati, UGM

Sri Noegrhay

Cause-effect (fish-bone) diagram of possible sources contributing to the uncertainty

30

Sri Noegrhay

Sri Noegrohati, UGM

Systematic error in sampling

31

The systematic effects in sampling are caused by the heterogeneity of the sampling target combined with an inability of the sampling method to properly reflect this heterogeneity. The heterogeneity can in turn be divided into the inherent heterogeneity of the material, caused by e.g. different size, shape and composition of the particles in a solid sample or different molecules in liquid samples, and distribution heterogeneity caused by e.g. poor mixing, which may allow particles or molecules of different characteristics to segregate in the target.

Sri Noegrhay

Sri Noegrohati, UGM

Reducing the systematic effects

32

Select methods for sampling and sample preparation that match the sampling target and its properties Increase the sample size give a better representation of the whole target Reduce the particle size of either the whole target or a relatively large sample, then collecting a sub-sample Mixing. This will reduce the segregation, However, in some special cases mixing may induce the segregation In these cases mixing should be avoided Proper storage or transportation should be carried out to reduce chemically and/or microbiologically changes of sample composition prior to the analysis

Sri Noegrhay

Sri Noegrohati, UGM

Random error in sampling

33

Random effects are mainly caused by variations in the composition of the sample in space or in time sampling method Sampling procedure or the handling of the sample, e.g. caused by different persons being involved The sampling equipment and the way in which the equipment works Approach to reduce the random effects is to increase the number of samples taken smaller standard deviation of the mean

Sri Noegrhay

Sri Noegrohati, UGM

The duplicate method: A balanced design

34

Appropriate ANOVA generates estimates of s2between-target , s2sampling , and s2analytical

The balanced design will only give the repeatability standard deviation of the analytical measurements, while the estimation of

analytical bias can be obtained from the well-matched certified reference materials Sri Noegrhay

Implications for planning sampling and measurement strategies.

35

1.

2.

3.

Expertise and consultation the sampling and analytical processes cover a range of activities all of those involved will have good knowledge of some part of the process, but few are able to advise on the complete process sample planners should involve analytical chemists, statistician, decision makers and other experts Avoiding sampling bias, include possible bias associated with differential sampling Planning for uncertainty estimation at least some replicated samples and measurements to assess the uncertainty of the results.
Sri Noegrhay

Implications for planning sampling and measurement strategies (contd)

36

4.

5.

6.

Fitness-for-purpose criteria, include the establishment of clear fitness-for-purpose criteria, taking into account the relative costs and uncertainties of sampling and analysis where they are known or can reasonably be determined in advance. Use of prior validation data, it should be noted that the variability observed during a relatively short series of analyses is rarely sufficient as an estimate of uncertainty. Long-term continuing validity studies are generally more reliable. Acceptability of sampling uncertainty should be evaluated
Sri Noegrhay

Competence requirements
37

To plan and perform qualified sampling and to make a reliable estimate of the measurement uncertainty require competence in the issue and the sampling target Theoretical and practical knowledge about the sampling method and the sampling equipment Sample analytical point of view e.g. stability, conservation, moisture uptake, how to avoid contamination and analyte loss etc. analytical method used, e.g. interferences, memory effects, sample amount needed, calibration strategy uncertainty in general

Sri Noegrhay

Sri Noegrohati, UGM

quality assurance of sampling

38

Sampling protocols are never perfect in that they can never describe the action required by the sampler for every possible eventuality that may arise in the real world in which sampling occurs. quality assurance of sampling, including The required competence, validation and quality control of sampling methods, documentation of sampling.

Sri Noegrhay

Sri Noegrohati, UGM

Illustration of the combined use of validation and quality control of sampling

39

One method used at many sites

Validation

One method used repeatedly at one site

Initial validation yielding On site validation yielding generic performance data the performance data for the specific target

Quality control

Extensive quality control with site specific verification of generic performance data

Spot quality control verifying the performance data consistency over time

Sri Noegrhay

Summary of sampling documentation

40

Sampling method A generic description of the operations used for sampling Sampling procedure A specific and detailed description of the operations used for sampling after a defined principle and with defined equipment. Sampling field report The detailed notes on the sampling details as noted in the field Chain of custody report A written record of the handling of the sample from sampling to analysis including transport and storage conditions. Sampling report Report summarizing the sampling results including target definition, reference to applied method and procedure, relevant notes from field and chain of custody report and uncertainty. Sri Noegrhay Sri Noegrohati, UGM

41

Sri Noegrhay

Sri Noegrohati, UGM

42

Sri Noegrhay

Sampling procedure and Sampling protocols

43

Sampling procedure ISO 3534-1: 4.5 (1993), ISO 11704-2, AMC (2005) Operational requirements and/or instructions relating to the use of a particular sampling plan; i.e. the planned method of selection, withdrawal and preparation of sample(s) from a lot to yield knowledge of the characteristic(s) of the lot. Sampling protocols describe the recommended procedure for the sampling of innumerable types of material and for many different chemical components. These protocols are sometimes specified in regulation or in international agreements (i.e. CAC/GL 33) These procedures rarely identify the relative contributions of sampling and chemical analysis to the combined uncertainty. Sri Noegrhay

Representative Samples
44

Sampling theory has developed largely independently of analytical chemistry and chemical metrology. Sampling quality has generally been addressed in sampling theory by the selection of a correct sampling protocol, appropriate validation, and training of sampling personnel (i.e. samplers) to ensure that the protocol is applied correctly It is then assumed that the samples will be representative and unbiased, and the variance will be that predicted by the model.

Sri Noegrhay

45

error and uncertainty in measurement

Sri Noegrhay

uncertainty of measurement
46

Uncertainty of measurement, or measurement uncertainty (MU), is defined in metrological terminology (ISO 1993) as: Parameter, associated with the result of a measurement, that characterises the dispersion of the values that could reasonably be attributed to the measurand.

The term value of the measurand is closely related to the traditional concept of true value in classical statistical terminology uncertainty has also been defined [ISO 35341: 1993] as:
An estimate attached to a test result which characterises the range of values within which the true value is asserted to lie Uncertainty is related to other concepts, such as accuracy, error, trueness, bias and precision.

Sri Noegrhay

47

Sri Noegrhay

48

The act of taking a sample introduces uncertainty into the reported measurement result wherever the objective of the measurement is defined in terms of the analyte concentration. Sampling protocols are never perfect . Sampling protocols never describe the action required by the sampler for every possible eventuality that may arise in the real world in which sampling occurs.

Heterogeneity always gives rise to uncertainty. If the test portion is a few micrograms nearly all material will be heterogeneous the sampling step will contribute to the uncertainty in the measurement of an analyte concentration processes of physical preparation (e.g. transportation, preservation, comminution, splitting, drying, sieving, homogenisation) introduce errors from a range of mechanisms, such as loss of analyte, loss of fine particles, or contamination from equipment or previous samples.
Sri Noegrhay

ERRORS
49

Presentation of Quantitative results should be accompanied by some estimate of the random (unpredictable) and systematic (predictable) errors in them. Random errors affect the precision of the result, systematic errors affect accuracy

Sampling plans are associated with two types of error: sampling error (caused by the sample failing to accurately represent the population from which it was collected); and measurement error (caused by the measured value of the characteristic failing to accurately represent the true value of the characteristic within the sample). the analysis should be quantified and minimised.
Sri Noegrhay

Systematic and random effects

50

The uncertainties caused by the sampling step can be divided systematic effects (bias) and random effects (precision), each being caused by a defined set of sources. Generally speaking the systematic effects are hard to quantify but often possible to avoid, whereas the random effects are easier to quantify but harder to avoid.
Sri Noegrhay Sri Noegrohati, UGM

Terms
51

Accuracy: The closeness of agreement between a test result and the accepted reference value related to systematic error. Bias: The difference between the expectation of the test result and an accepted reference value. Note: Bias is a measure of the total systematic error as contrasted to random error. Precision: The closeness of agreement between independent test results obtained under stipulated conditions. Notes: 1. Precision depends only on the distribution of random errors 2. The measure of precision usually is expressed in terms of imprecision and computed as a standard deviation of the test results. 3. Independent test results means results obtained in a manner not influenced by any previous result on the same or similar test object Sri Noegrhay

Sources of sampling uncertainty in sampling theory of Gy 1992

Source Fundamental sampling error (FSE) Description A result of the constitutional heterogeneity (the particles being chemically or physically different) A result of the distributional heterogeneity Trends across space or over time Periodic levels across space or over time Identifying the correct sample to take. Considers the volume boundaries of a correct sampling device Removing the intended sample. Considers the shape of the sampling device cutting edges Contamination (extraneous material in sample): Losses (adsorption, condensation, precipitation etc.): Alteration of chemical composition (preservation): Alteration of physical composition (agglomeration, breaking of particles, moisture etc.):

Grouping and segregation error (GSE) Long-range point selection error (PSE1) Periodic point selection error (PSE2) Increment delimitation error (IDE)

Increment extraction error (IXE) Increment and sample preparation error (IPE)

Weighting error (SWE)

52

The result of errors in assigning weights to different parts of an unequal composite sample

Sri Noegrhay

53

Sri Noegrhay

Fitness For Purpose

54

Fitness for purpose is the degree to which data produced by a measurement process enables a user to make technically and administratively correct decisions for a stated purpose. The fitness for purpose of measurement results can only be judged by having reliable estimates of their uncertainty. Require an effective procedures for estimating the uncertainties arising from all parts of the measurement process, includes uncertainties arising from any relevant sampling and physical preparation. understanding uncertainty in sampling provides procedures that allow their practical implementation

Sri Noegrhay

Approaches to uncertainty estimation

55

Both sampling and analysis contribute to measurement uncertainty. There are two main approaches to the estimation of uncertainty The empirical approach (top-down ) uses repeated sampling and analysis, under various conditions, to quantify the effects caused by factors such as the (1) heterogeneity of the analyte in the sampling target, (2) variations in the application of sampling protocols The modelling approach (bottom-up) uses a predefined model that identifies each of the component parts of the uncertainty, making estimates of each component, and sums them in order to make an overall estimate.
Sri Noegrhay

Empirical approach (top-down)

56

intended to obtain a reliable estimate of the uncertainty, without necessarily knowing any of the sources individually. It relies on overall reproducibility estimates from either in-house or interorganisational measurement trials. It is possible to describe the general type of source, such as random or systematic effects, and to subdivide these as those arising from the sampling process or the analytical process. Estimates of the magnitude of each properties of the measurement methods separately sampling precision (for random effects arising from sampling) analytical bias (for systematic effects arising from chemical analysis These estimates can be combined to produce an estimate of the uncertainty in the measurement result. Sri Noegrhay

measured and true values

57

the relationship between the measured single measurement of analyte concentration (x), on one sample (composite or single), from one particular sampling target and true values of analyte concentration :

Xtrue is the true value of the analyte concentration in the sampling target, The total error due to sampling is sampling and the total analytical error is analysis if the sources of variation are independent, the measurement variance

If statistical estimates of variance (s2) are used

Sri Noegrhay

Estimating Standard uncertainty

58

Both sampling and analysis contribute to measurement uncertainty The standard deviation of the measurement

The ssampling can then be obtained by

The random part of the uncertainty is described by the standard deviation The standard uncertainty (u) can be estimated using smeas

Sri Noegrhay

Sri Noegrohati, UGM

Concentration variation between the targets

59

In a survey across several sampling targets recommend the additional term target represents the variation of concentration between the targets and has variance 2between-target The total variance 2total

replicate method

Sri Noegrhay

Combined Standard Uncertainty (u)

60

The combined standard uncertainty, u, is calculated based on standard deviations of replicate measurements, x. It may, or may not, include contributions from systematic effects to get the combined standard uncertainty (u) of sampling and analysis, estimates of systematic effects should be included. the systematic errors (bias) cannot be easily obtained when duplicate design is used, but some approaches to this are given The bias of analysis can be estimated by using certified reference materials (CRM) or participating in laboratory proficiency tests.

Sri Noegrhay

Expanded Uncertainty (U)

61

The combined standard uncertainty, u, is calculated based on standard deviations of replicate measurements, x. It may, or may not, include contributions from systematic effects In any case, as it is based on one single standard deviation ( X = x u ) will mean that the probability that the reported range contains the "true value" is only 67% (a 67% confidence interval). In most cases, it is therefore more useful to the persons evaluating the data to use the expanded uncertainty, U the expanded uncertainty, U, obtained from replicate measurements, x, applying a coverage factor of 2 U = 2 u the result, x, reported as X = x U , giving the range of the true value, X, with 95% confidence.
Sri Noegrhay Sri Noegrohati, UGM

Calculation of uncertainty and its components

62

The values of ssamp and sanal from the ANOVA are estimates of sampling precision and analytical precision respectively. The random component of the measurement uncertainty is calculated by the combination of these two estimates The expanded uncertainty, for approximately 95% confidence for example, requires this value to be multiplied by a coverage factor of 2. U can also be expressed relative to the reported value x and expressed in terms of a percentage,

Sri Noegrhay

relative expanded uncertainty

63

The relative expanded uncertainty for the sampling or analysis alone can similarly be expressed as

Sri Noegrhay

Examples of tools for the estimation of uncertainty contributions

64

Random (precision)
Analysis Replicate analyses

Systematic (bias)
Certified reference materials Laboratory proficiency test Reference analytical method Reference sampling target Sampler proficiency test, Inter-method comparisons Known theoretical value of sampling target Reference sampling method

Sampling

Replicate samples

Sri Noegrhay

Sri Noegrohati, UGM

65

The modelling approach (bottom-up)

Initially, identifies all of the sources of uncertainty, such as the form of a cause-and-effect (fishbone), quantifies the contributions from each source, and then combines all of the contributions to give an estimate of the combined standard uncertainty. The uncertainty of measurement generated by each of these steps is estimated independently, then calculated by combining the uncertainty from all of the steps by appropriate methods. This approach is well established for analytical Sri Noegrhay methods [1]

Sampling theory for estimation of uncertainty

66

This approach relies on the use of a theoretical model (Pierre Gy), Most sampling errors, except the preparation errors, are due to the material heterogeneity, which can be divided into two classes: 1) constitution heterogeneity (CH), and 2) distribution heterogeneity (DH). Both heterogeneities can be mathematically defined and experimentally estimated.

Sri Noegrhay

total standard deviation

67

The total standard deviation is given by the formula: where s is the sampling standard-deviation, m the measurement standard-deviation

- the most frequent one: the analytical error << the sampling error, i.e the analytical error is at most equal to one third of the sampling error In this case, m s/3

The standard deviation for the observed results will be at most 5 % larger than the sampling standard deviation taking into account the analytical error. Sri Noegrhay

68

Principles of quality assurance in sampling

Sri Noegrhay

Relationship between validation and quality control

69

After the required uncertainty which makes the measurements fit for purpose has been established the sampling and analytical procedures proposed to meet those purposes should be evaluated tools required are: validation and continuous quality control..

Sampling validation comprises a one-time estimation determined under conditions expected in the routine use of the sampling procedure demonstrates what can be achieved but not yet shows the conformation to fitness-for-purpose requirements For sampling, where the degree of heterogeneity may vary markedly from one target to the next the larger part of the uncertainty component stems from the heterogeneity of the target validation alone cannot ensure that routine results are indeed fit for purpose Sri Noegrhay Sri Noegrohati, UGM

Methods of internal quality control of sampling

70

we need to see whether results for individual sampling targets are fit for purpose, Bias is difficult to address in validation and almost impossible in internal quality control. The focus of interest is the precision aspect The principal tool is replication minimally executed by taking two samples from each target by a complete (and suitably randomised) duplication of the sampling protocol Each sample is analyzed once. the difference between the results

the standard deviation of measurement

Sri Noegrhay

71

If the validated uncertainties of sampling and analysis are us and ua respectively, the combined standard uncertainty is

a one sided range control chart can be constructed with a control limit (at the 95% confidence interval) of 2.83umeas and an action limit (at the 99% confidence interval) of 3.69umeas

Sri Noegrhay

Example of a range control chart for quality control of sampling

72

Central line: CL = 1.128* smeasurement Warning limit: WL = 2.83* smeasurement (not exceeded in 95% of control result) Action limit: AL = 3.69* smeasurement (not exceeded in 99% of control result)

Sri Noegrhay

Judging fitness for purpose of measurements using uncertainty

73

A proper understanding of uncertainty from sampling must be embedded in the broader perspective of fitness for purpose. Three approaches have been suggested for setting fitness-forpurpose criteria. First : to set an arbitrary limit on the maximum value of uncertainty that is considered acceptable. does not relate to intended purpose for which the user requires the measurement. Second: to compare the variance generated by the measurement (sampling and analysis) to the variance of the different sampling targets, such as in mineral exploration sets the fitness-for-purpose criterion so that the measurement variance < 20% to the total variance third, to judge the fitness for purpose of measurements consider the effect of the measurement support a decision. A decision can be either correct or incorrect, an incorrect decision Sri Noegrhay is more likely if the uncertainty is higher.

74

need the quality requirements for sampling and the uncertainty associated with between target variability. If the uncertainty of measurements is underestimated, i.e. sampling uncertainty is not taken into account erroneous decisions may be made large financial, health and environmental consequences.

Sri Noegrhay

Sri Noegrohati, UGM