The sign test is a nonparametric statistical test based only on whether observations lie above or below a hypothesized statistical
median. For paired data, it is applied to the signs of the nonzero paired differences.
Under a continuous null distribution with median
zero, the number 
of positive differences among 
nonzero differences has the binomial distribution
 |
The P-value is therefore computed from binomial tail probabilities. Zero differences require a specified convention and are commonly omitted. The test uses less information than the Wilcoxon signed rank test but requires no symmetry assumption.
