Testing �d
We routinely test to see whether the sample mean difference is significantly different from 0 ("no expected difference"). A
formal 4 step procedure is recommened. Steps are:
- (A) Statement of the null and alternative
- The two sided null and alternative are H0: � = 0 vs. H1: � not = 0
- (B) Set alpha level
- Alpha = something (usually .05 by convention)
- (C) Calculate test statistics
- Paired t-statistic = (sample mean dif) / (sem), df = n - 1, the two-sided p value corresponds to the tail probabilities
beyond the absolute value of the t-statistic on Students t with df degrees of freedom
- These statistics are calculated automatically when Epi Info one-variable MEANS command is properly directed
toward the DELTA variable
- There are also several Web calculators that can be used to "crunch" the numbers
- (D) State conclusion
- The null hypothesis is rejected if p < alpha. Otherwise, the null hypothesis is retained for want of evidence.