Universal kriging originally was developed for problems of spatial int
erpolation if a drift seemed to be justified to model the experimental
data. Bur its use has been questioned in relation to the bins of the
estimated underlying variogram (variogram of the residuals), and furth
ermore universal kriging came to be considered on old-fashioned method
after the theory of intrinsic random functions was developed. In this
paper the model is reexamined together with methods for handling prob
lems in the inference of parameters. The efficiency of the inference o
f covariance parameters is shown in terms of bias, variance, and mean
square error of the sampling distribution obtained by Monte Carlo simu
lation for three different estimators (maximum likelihood, bias correc
ted maximum likelihood, and restricted maximum likelihood). ii is show
n that unbiased estimates for the covariance parameters may be obtaine
d but if the number of samples is small there can be no guarantee of '
good' estimates (estimates close to the true value) because the sampli
ng variance usually is large. This problem is not specific to the univ
ersal kriging model but rather arises in any model where parameters ar
e inferred from experimental data. The validity of the estimates may b
e evaluated statistically as a risk function as is shown in this paper
.