Loading...
Please wait, while we are loading the content...
Similar Documents
How to construct a confidence interval fromonlyonemeasurementonacomposite sample assuming log-normality and known variance for the increment samples
| Content Provider | Semantic Scholar |
|---|---|
| Author | Voet, H. Van Der |
| Copyright Year | 2005 |
| Abstract | You have to decide whether a lot of animal feed can be accepted on the basis of one analysis result. You know that the material is non-homogeneous, so that sampling uncertainty should be considered. There is some historical knowledge about the variability of the material: How to proceed? For inspection purposes a lot is often sampled according to relevant regulations (e.g. EU directives 76/371 [1] or 98/53 [2]), by combining a number of increment samples into one composite (bulk or aggregate) sample, from which subsequently only one measurement of a characteristic of interest is obtained. An example is the official inspection of aflatoxin levels in lots of animal feed products. For the purpose of comparing the measurement result with a regulatory limit value it is often necessary to specify the sampling uncertainty of the measurement result. In this note we will assume that sampling uncertainty overwhelms measurement uncertainty. Clearly, sampling uncertainty cannot be derived from one single value, and therefore we assume that knowledge is available about the heterogeneity of the characteristic from other, similar lots of material. For many characteristics of interest, such as low-level chemical residue concentrations, the lognormal distribution is a sensible model to describe the distribution across the lot. The problem is that the measurement on the composite sample is no longer a direct observation from this lognormal distribution, but is the arithmetic mean of several observations from the same lognormal distribution. For the situation where several composite samples are taken (from the same lot), it has been described how to use both mean and variance of the set of measurements to construct a confidence interval for the mean [3, 4]. However, for the much simpler situation of only one composite sample, but where the knowledge about variability of the lot is assumed to be known, no explicit expression of the confidence interval was found. In this note we derive this expression in a form suitable for practical use. |
| Starting Page | 452 |
| Ending Page | 454 |
| Page Count | 3 |
| File Format | PDF HTM / HTML |
| DOI | 10.1007/s00769-005-0013-8 |
| Volume Number | 10 |
| Alternate Webpage(s) | http://edepot.wur.nl/40321 |
| Alternate Webpage(s) | https://doi.org/10.1007/s00769-005-0013-8 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
