Part I
Part II
Part III
|
FAQ - Coal Analyses (Part II)
1. What are the specifics surrounding the statistical confidence interval?
On pages 24 and 25 of the Supporting Statement, EPA proposed a statistical test at the end of each quarter to determine future coal sampling requirements. The Agency, however, noted that:
-
Sufficient data were unavailable to determine whether or not a +/- 10 percent of a 90
percent confidence interval about the mean amount of mercury contained within the coal
was attainable. If data become available before reporting begins on January 1, 1999
that indicates this percentage should be higher or lower, proper adjustments will be
made.
The Electric Power Research Institute (EPRI) using historical mercury in coal data presented in the Electric Power Research Institute's report Mercury and Other Trace Metals in Coal, TR-106950, January 1997 demonstrated to the EPA that almost every power utility would fail the statistical criterion of 10% tolerance (i.e., that the lower confidence limit and upper confidence limit must be within 20% of the sample average). Based on further evaluation of this coal data, the EPA has determined that increasing the tolerance from 10% to 30% would provide a fair evaluation, while maintaining mercury in coal data acceptable for making a regulatory determination on mercury emissions from power plants. Therefore, the statistical test that should be used to determine quarterly sampling requirements is that the lower and upper 90% confidence limits fall within 30% of the sample average.
- Doesn't the alpha(0.05) correspond to a 95 percent confidence interval?
Answer: An alpha of 0.05 corresponds to a one-tailed confidence limit (i.e., lower confidence limit or upper confidence limit, not both) of 95 percent or a two-tailed confidence interval (i.e., an upper and lower confidence limit) of 90 percent.
- Shouldn't it be alpha (0.1) for a 90 percent confidence interval?
Answer: No. An alpha of 0.1 would correspond to a 90 percent confidence limit (i.e., lower confidence limit or upper confidence limit, not both).
- Is alpha one or two tailed?
Answer: Alpha is two-tailed.
- Can the t-values be taken from critical tables or does EPA have a specific set of t-values to use?
Answer: The t-values can be taken from the critical tables. Below is the table of t-values for anyone who does not have it. Only the Excel and Lotus files should be used in creating spreadsheet macros to perform the statistical analysis automatically.
- Does alpha use n or (n-1)?
Answer: Alpha uses n (i.e., t0.05 is a function of n).
Here is an example of the statistical calculation:
As of April 15, 1999, the calculation can be performed by following the Statistical Evaluation link on the web application's main menu.
OMB Control Number 2060-0396
Last Update:
27 April 1999
utilityhelp@rti.org
|
