Site-to-site comparisons

ISM data from one site can be compared to that from another. For example, sampling for two DUs may consist of 30 increments and 5 replicates each. Standard two-sample hypothesis tests (e.g., Wilcoxon Rank Sum or Gehan) can be applied to determine whether the differences are statistically significant under the assumption that the variances are the same. While similar in concept to distribution testing with discrete sampling data, some aspects of ISM data comparisons are unique. First, while it is not necessary for the DUs to have the same number of ISM replicates, it is likely that the number of replicates will be quite small (e.g., 35). Therefore, the decision errors may be higher with hypothesis testing using ISM data compared to discrete data. In addition, since the estimated SD of ISM replicate results is a function of the number of increments obtained from the DU, samples can be compared directly only if the same number of increments is collected. To conduct a hypothesis test for ISM data based on unequal increments, a statistical adjustment to the estimated SD may be appropriate to reduce the chance of violating the hypothesis test assumption of equal variance (see Appendix A).