[ Pobierz całość w formacie PDF ]
.16.5 The optimal apportionment of resources between sampling and analysis is also amatter of costs.Even an elementary consideration (excluding costs) shows that theuncertainties of sampling and analysis should be roughly balanced.For example, if theuncertainties of sampling and analysis are 10 and 3 units respectively, the overall uncertaintyof measurement is 102 + 32 = 10.4.The overall uncertainty is hardly affected by a reductionof the uncertainty of analysis: if it is reduced to (say) 1 unit, the overall uncertainty is reducedto 102 +12 = 10.05 , an inconsequential change.A more sophisticated approach takes intoaccount the different costs of analysis and sampling.If the unit costs of sampling and analysisare A and B for the same specific level of uncertainty, the optimum ratio of samplinguncertainty usamp to analytical uncertainty uanal is given byusamp # A #1 4=.# #uanal # B#2 2This ratio provides the minimum expenditure for a given overall uncertainty of usamp + uanalor, alternatively, the minimum uncertainty for a given expenditure [34].Methods for modifying uncertainty from sampling are discussed in Appendix E, althoughoperating at minimum total cost is not always achievable or necessary.UfS:2007 Page 31 Implications for planning sampling and measurement strategies17 Implications for planning sampling and measurement strategies17.1 Expertise and consultationAs Section 4 shows, the sampling and analytical processes cover a range of activities.Different parts of the process are frequently allocated to different staff, who may have verydifferent knowledge of the objectives and, more importantly, differing knowledge of theeffect of different parts of the process.In general, all of those involved will have goodknowledge of some part of the process, but few are able to advise on the complete process.Itis therefore important that sample planners involve analytical chemists and experiencedsampling technicians where possible in planning sampling.It is also prudent to includestatistical experts in most circumstances (see below).Decision makers (i.e.business managersand those acting on the results of sampling activities) should be involved in planning for newapplications, and regulators should also be consulted where a protocol is intended to supportregulation.Although the principles of this Guide are widely applicable, expert statistical guidance isalways valuable and should be considered essential in some circumstances.These include:" where the observed or expected frequency distributions are not normal, for example wherethe results contain more than 10% outliers, or where the results show markedlyasymmetric distributions;" where large financial or social consequences depend on a reliable estimate of uncertainty;" where confidence intervals are needed on the estimates of uncertainty or, for morecomplex sampling plans, on the measurement results;" where the sampling strategy is more complex than simple random sampling withreplicated measurements, for example in implementing stratified sampling.17.2 Avoiding sampling biasThe methods described in this Guide are suitable for establishing the variability of sampling,but only the more complex methods can begin to assess uncertainties associated with possiblebias in sampling.For this reason, close attention should be paid to minimising potentialsources of bias.These include possible bias associated with differential sampling due toparticle size, density or flow-rate; bias in selection of sampling points; the effect of differentsampling equipment etc.Specific expertise in sampling methodology should be sought unlessthese factors can be demonstrated to be adequately controlled or are completely specified byan established sampling protocol.17.3 Planning for uncertainty estimationSampling exercises should always make provision for at least some replicated samples andmeasurements in order to assess the uncertainty of the results.17.4 Fitness-for-purpose criteriaPlanning should ideally begin with the establishment of clear fitness-for-purpose criteria,taking into account the relative costs and uncertainties of sampling and analysis where theyare known or can reasonably be determined in advance.Section 16 provides guidance on howanalytical and sampling effort can be optimised.UfS:2007 Page 32 Implications for planning sampling and measurement strategies17.5 Use of prior validation dataThe main uncertainties associated with analytical measurements are often estimated during, oron the basis of, analytical method validation, a process which is carried out prior to bringingthe method into use.Consideration accordingly needs to be given as to whether the variabilityfound as part of the sampling experiment should replace, inform, or simply serve as a checkon, the analytical measurement uncertainty assessed using prior information.In consideringthis issue, it should be noted that the variability observed during a relatively short series ofanalyses is rarely sufficient as an estimate of uncertainty.Long-term studies are generallymore reliable.It is accordingly safer to rely on prior validation data unless the observedvariation is significantly higher.Uncertainties associated with sampling variability can themselves be estimated in advance,particularly where a long-term sampling programme is to be planned and implemented.Underthese circumstances, it is usually prudent to obtain an initial estimate of sampling uncertainty
[ Pobierz całość w formacie PDF ]