|
In the estimation process of long-term noise indicators
Ld, Lw, Ln, Ldwn there is the problem of selecting the measurement sample size. On the one hand, increasing the size
of the measurement sample is beneficial in terms of obtaining uncertainty ranges for noise indicators with the desired
statistical properties – adequate coverage and length. On
the other hand, it is economically unjustified. Most of the
classical algorithms for determining the measurement uncertainty are based on the assumptions of the Central Limit
Theorem, allowing to assign average energy levels or average sound levels of normal distribution. Such assumptions
enable the construction of uncertainty intervals. When using the CLT the minimum sample size is not strictly defined. Although there are formulas that allow to determine
the minimum sample size so that the CLT can be used
(e.g. Cochran’s formula). However, this formula does not
take into account, for example, the problem of rounding
measurement data and their accuracy. It may be the case
that the minimum sample size calculated from Cochran’s
formula is several dozen elements whereas for a dozen or
so measurement samples the ends of the uncertainty intervals do not differ significantly (as to the magnitude of the
measurement error) from the range determined by classical
methods. The article presents the problem of determining
the minimum size of the measurement sample of long-term
noise indicators. It is also presented how the uncertainty
intervals.
|
