3. Data analysis
All perceived (from the questionnaire interviews) and measured
data were analyzed using the Statistical Package for Social Sciences
(SPSS) version 16 (SPSS,2007). Statistical variationof categorical data
(perception data other than cattle productivity)was tested bymeans
of cross tabulation, assuming significant differences when P < 0.05,
whereas post hoc multiple mean comparisons were employed for
numerical data (cattle productivity as well as measured rangeland
ecological conditions) in one-wayANOVAusing the Duncan multiple
range test and assuming significant differences when P < 0.05.
Descriptive statistics, such as means, percentages and standard error
of the means were used to present the results of perceptions and
measurements. Further, the relationship between cattle productivity
and measured botanical and soil parameters was analyzed by
Spearman’s rho correlation. A series of stepwise linear regression
analyses were conducted with soil erosion as the independent variable
and other measured and numerical perception variables as dependent variables across the altitudes.