Dr. T.J. LitwinskaKemperink 

Geostatistical techniques have become a
standard part of commercial
interpretation software. However, these
methods should not be regarded as a black
box, disconnected from the constraints
provided by the statistical theory. 
Consequently:
 empirical data must be tested for stationarity and stochastical
independence;
 sufficient big sample size or a normal distribution of measured data may
be a further requirement to be proved statistically;
 significance tests and confidence intervals are important.
Therefore, complex statistical calculations, such as:
 analysis of variance and regression
 analysis of trend
 Kriging and CoKriging
should be performed individually, problemoriented and depending on data
quality.
Probability estimation based on semiMARKOV models
Different log data of bore holes penetrating cyclic sedimentation are
investigated.
The estimated probability matrix describes the transitions between different
lithological units, whereas the derived conditional distribution functions give
the distribution of thickness.
A semiMARKOV model is useful for:
 Comparison of different sedimentation environments;
 Simulation of sedimentation processes;
 Log reconstruction. 





