I have tried different geo-statistical interpolation technique to evaluate the interpolation results of a linearly distribution data but, kriging, idw and other methods have not yielded good performance in terms of statistical significance (e.g R2, RMSE value). I know that the geostatistical module of ArCGIS uses Leave One Out Method for Cross-Validation but, is there any other option to carry out K-fold cross validation within GIS platform? How can I validate my interpolation results showing which would show high R2, RMSE values from scratch? Can we go through an iterative process to randomly select all 2/3 data for interpolation and 1/3 data for validation to get the best model?