Appendix C. Indicators of Hydrologic Alteration Results

The output results from the Indicators of Hydrologic Alteration software are included in a set of Microsoft^© Excel files, which can be downloaded from http://pubs.usgs.gov/sir/2010/5016.

In the directory there are 12 Excel files, which pertain to the study Reaches 1–12 (for example, R1obs.xls).

Statistical Testing:

The Indicators of Hydrologic Alteration software includes a “significance count” as a means of testing if the difference between a pre- and post-impact period metric is significant. The significance count value can be interpreted similarly to a p-value. A description of the test from the software manual (p. 47) is shown below:

“Columns 7 and 8 calculate a “significance count” for the deviation values. To calculate this, the software program randomly shuffles all years of input data and recalculates (fictitious) pre- and post-impact medians and CDs 1000 times. The significance count is the fraction of trials for which the deviation values for the medians or CDs were greater than for the real case. So a low significance count (minimum value is 0) means that the difference between the pre- and post-impact periods is highly significant, and a high significance count (maximum value is 1) means that there is little difference between the pre- and post-impact periods. The significance count can be interpreted similarly to a p-value in parametric statistics.”

“It is important to understand that in some infrequent situations this algorithm may generate very low significance counts when there is very little apparent difference between the pre- and post-impact periods. This can occur when the deviation factor between the pre- and post-impact periods is zero or very small, and the overall distribution contains a large number of values right at or very near the center of the distribution. In this situation a low significance count actually means that the lack of difference between the two periods is highly significant, in a statistical sense, because randomly rearranging the data rarely yield a larger deviation factor than the original data.”

“It also is important to understand that significance counts may differ slightly each time the IHA is run for the same dataset, since a new set of randomized cases is generated each time.”

First posted February 8, 2010