1: \begin{abstract}

2: Information criteria, such as Akaike's information criterion and

3: Bayesian information criterion are often applied in model selection.

4: However, their asymptotic behaviors for selecting geostatistical

5: regression models have not been well studied, particularly under the

6: fixed domain asymptotic framework with more and more data observed

7: in a bounded fixed region. In this article, we study the generalized

8: information criterion (GIC) for selecting geostatistical regression

9: models under a more general mixed domain asymptotic framework.

10: Via uniform convergence developments of some statistics,

11: we establish the selection consistency and the asymptotic loss efficiency of GIC

12: under some regularity conditions, regardless of whether the covariance model is correctly

13: or wrongly

14: specified. We further provide specific examples

15: with different types of explanatory variables that satisfy the conditions. For example,

16: in some situations, GIC is selection consistent, even when some

17: spatial covariance parameters cannot be estimated consistently.

18: On the other hand, GIC fails to

19: select the true polynomial order consistently under the fixed domain

20: asymptotic framework. Moreover, the growth rate of the domain and

21: the degree of smoothness of candidate regressors in space are shown to play key

22: roles for model selection.

23: \end{abstract}