Understand what you are doing and then try different models.

First you set rigid physical models in line with known parameters. So the correlation you are setting up is minimum and maximum values for the physical aerodynamic attributes (such as max warp dimension of wings, air resistance values, pressure/lift differentials). The more set/valid values associated with the spatial points of the model the more accurate the model. The regression analysis is interpolating between these set points.

Second, if not using standardized Kriging routine from a publication (literature should contain values used in analysis) or if unsure of the values to use when setting the model- play! That's what I do. Once set in ArcGIS you can apply a range of different Kriging routines with variations in individual settings for each run. Look for the one that 'looks' realistic from the range of test runs.