Comparison of normality test methods for some soil properties in the arid land of South Khorasan

Abstract

Statistical assumptions are the basis of many univariate and multivariate statistical tests. Normality is the most basic assumption of multivariate analysis in plant ecology. If the normality assumption is violated, some specific statistical tests are not valid. Therefore, the present study compares the methods of normality assessment of some soil properties in the arid land of South Khorasan. It also examines the effect of increasing the number of soil samples from 25 to 50 on the normality results. Histogram, box plot, Q-Q plot, CV, skewness, and univariate and multivariate normality tests were used. The results showed that EC, K, Ca, Mg, Na, Cl, HCO3, and SAR data had a very high variation (CV 75–100%) and saturation moisture and pH had a low variation (CV <15%). Based on the results of most statistical tests and the skewness coefficient, saturation moisture, pH, N, P, CaCO3, sand, and silt were normal. EC, K, Ca, Mg, Na, Cl, HCO3 and SAR had the right skewed distribution. The results showed multivariate normality was violated, and the use of these data was not suitable for multivariate analysis. The results of the goodness-of-fit test showed that P, sand, and silt follow a normal distribution. Other soil properties do not follow any of the studied probability distributions (p≥0.05). Therefore, the use of nonparametric is recommended for the physical and chemical properties of the soil in the area. Although in general, the increase in the number of samples has a positive effect on the actual distribution of the community, but due to the high spatial variability of some soil properties such as salinity, the status of nutrients, particle size, etc., the CV and the range of variations in most of soil properties are wide.