VDI 2448 BLATT 2
Statistical evaluation of random-sample measurements of stationary source emissions: Determination of the upper confidence limit
Organization: | VDI |
Publication Date: | 1 July 1997 |
Status: | active |
Page Count: | 14 |
ICS Code (Air quality in general): | 13.040.01 |
scope:
Introduction
Random-sample emission measurements are carried out in Order to obtain, with minimum measurement expense, the information required to answer a previously formulated measurement task. It is not possible in principle to specify exactly a true characteristic (such as an average value, percentile or maximum value).
However, on the basis of random-sample measure¬ ments it is possible in principle to estimate characteristics for the totality of the emitted concentrations.
Calculating simple numbers as an estimate of the characteristic quantities is not sufficient to be able to make reliable statements. On the contrary, it is necessary, starting from the type of characteristic to be estimated, from the sample size and from the actual data in the random sample, to specify an interval (confidence interval) which contains the true characteristic with a prescribed Statistical confidence (confidence level). The upper confidence limit specifies the limit below which the true value lies for a preselected Sta¬ tistical confidence level.
Various parameters feature in the upper confidence limit, such as random measuring errors and the sample size, The value of the upper confidence limit also depends, however, on the choice of the characteristic under review (average value, percentile, maximum value, etc,).
The following characteristics and the confidence in statements related to their use are treated in this guideline:
- arithmetic average value
- percentile value
- maximum value
This guideline contains a tabular annex with calculating factors for determining the upper confidence limit of small random-sample populations (N < 10). If necessary, the measurement series can be supplemented later with further measured data. The measured data can come from various measurement series. A common evaluation is possible only given comparable boundary conditions.