Quantile Regression: Salmonid Abundance vs. Metals Toxicity Units for Colorado Streams

Plot

1. Analysis: Quantile regression
2. Independent variable: Metal Toxicity Units
3. Dependent variable: Number of Salmonids
4. Key: dot=observation; line= 90^th quantile
5. Sample size: n = 30

Model

1. Number of Salmonids = 153.3 - 111.364*(Metal Toxicity Units)
2. Date: 12 November, 2006

Variable	Estimate	lower CL	upper CL	Test Type	Statistic	p
Intercept	153.300	-178.102	484.702	T value	0.948
Metal Toxicity Units	-111.364	-656.511	433.782	T value	-0.418	0.679
Wald				Chi2	0.175	0.676
Likelihood Ratio				Chi2	6.494	0.011

Data Analysis

1. Model generated using SAS Proc QUANTREG with the smoothing algorithm. The WALD and Likelihood test statistics were used to test the linear null hypothesis, H₀: slope=0.
2. Analysis for sites classified as pH neutral, but with strong AMD input.
3. Metal Toxicity Units (TU) were calculated as sum of the ratios of each measured metal concentration with the criterion for that metal.
4. Replicate visits to a site were not excluded from the data set.

Data Origin

1. Agency: U.S. Environmental Protection Agency
2. Program: Environmental Monitoring and Assessment Program (EMAP)
3. Project: Colorado Metals Belt Stream Assessment (EPA Region 8)
4. Filenames: Files benmet, fishmet and chem downloaded from the Surface Waters Information Management (S.W.I.M.S.) database web site.

Sampling Design

1. Data collected 1994 through 1995.
2. Analyzed metals and metalloids: aluminum, arsenic, cadmium, chromium, copper, iron, manganese, nickel, lead, selenium, silver, zinc.
3. A description of the sampling methods may be found in the EMAP West Methods Manual for Streams found on the S.W.I.M.S. database web site.

Contact

Patricia Shaw-Allen, shaw-allen.patricia@epa.gov, (513) 569-7136