Accounting For Preferential Sampling In Geostatistical Inference