A statistical analysis was made of the activity concentrations measured in surface and deep soil samples for natural and anthropogenic g-emitting radionuclides. Soil samples were obtained from 48 different locations in Gilgit, Pakistan covering about 50 km2 areas at an average altitude of 1550 m above sea level. From each location two samples were collected: one from the top soil (2-6 cm) and another from a depth of 6-10 cm. Four radionuclides including226Ra, 232Th 40K and 137Cs were quantified. The data was analyzed using t-test to find out activity concentration difference between the surface and depth samples. At the surface, the median activity concentrations were 23.7, 29.1, 4.6 and 115 Bq kg-1 for 226Ra, 232Th, 137Cs and 40K respectively. For the same radionuclides, the activity concentrations were respectively 25.5, 26.2, 2.9 and 191 Bq kg-1 for the depth samples. Principal component analysis (PCA) was applied to explore patterns within the data. A positive significant correlation was observed between the radionuclides 226Ra and 232Th. The data from PCA was further utilized in linear discriminant analysis (LDA) for the classification of surface and depth samples. LDA classified surface and depth samples with good predictability.