Bayesian soil assessments combining prior with posterior censored samples

The geotechnical engineer is often faced with the problem of how to assess the statistical properties of a soil parameter on the basis of a sample measured in-situ or in the laboratory with the defect that some values have been replaced by interval bounds because the corresponding soil parameter values have turned out to exceed the capacity of the measuring instrument.

Given such a censored sample the problem is to estimate the mean value, variance and certain fractiles of the distribution of the soil parameter as it would be measured by an instrument of unrestricted range. In many situations only a small censored sample is given. To estimate the characteristic value defined as a lower fractile value corresponding to a codified probability value, the geotechnical engineer is thus urged to supplement the information from the measurements at the actual location by considering whatever prior knowledge is available about the soil parameter distribution.

The present paper shows how a characteristic value by computer calculations can be assessed systematically from the actual sample of censored data supplemented with prior information from a soil parameter data base.