Uncertainty and Statistical Risk

Abstract

All measurements are subject to uncertainty and the measurement result is complete only when accompanied by a statement of uncertainty associated with the measured value. This is the case for most statistical indicators, from the conjunctural predictions, to the measurement of household’s expenditure and income, to the calculation of some GDP components, to determining of the price indexes, to determining the unobservable economy, to determine the population’s perception regarding quality of life etc., up to voting intentions, to list only a few areas. The uncertainty, in the field of measurement, must be materialized by a statistical indicator, which expresses a certain fact, the distance /closeness to the true value of the size subject to the measurement process. Uncertainty also appears as the result of human ignorance, and its form of manifestation is the variability which, exceeding certain admissible limits, can generate what we commonly call a risk, namely to make an erroneous decision in a situation where necessary information is distorted precisely because of too much variability. The risk in making decisions is present in all human activities, from where the vastness of the problem as a research field. In the paper we propose the exposure of uncertainty measurement procedures and the link between uncertainty and risk. The statistical modeling of risk has as a starting point the assumption that risk can be assimilated with the possibility of suffering a certain loss. Because the possibility is expressed quantitatively, by probability, the risk appears as a probability function in the occurrence of an unwanted phenomenon. Finally, certain Taguchi risk applications are presented, regarding the relationship between this risk and the potential index of a process.

Keywords

• uncertainty
• errors of iii and iv type
• statistical risk
• taguchi method