[Federal Register Volume 62, Number 171 (Thursday, September 4, 1997)]
[Rules and Regulations]
[Pages 46679-46680]
From the Federal Register Online via the Government Publishing Office [www.gpo.gov]
[FR Doc No: 97-23404]
-----------------------------------------------------------------------
DEPARTMENT OF COMMERCE
National Oceanic and Atmospheric Administration
50 CFR Part 622
[Docket No. 961226370-7074-02; I.D. 111896A]
RIN 0648-AI15
Fisheries of the Caribbean, Gulf of Mexico, and South Atlantic;
Shrimp Fishery Off the Southern Atlantic States; Amendment 2;
Correction
AGENCY: National Marine Fisheries Service (NMFS), National Oceanic and
Atmospheric Administration (NOAA), Commerce.
ACTION: Final rule; correction.
-----------------------------------------------------------------------
SUMMARY: NMFS published in the Federal Register of April 16, 1997, the
testing protocol for bycatch reduction devices (BRDs). The statistical
approach section of that appendix contains errors. This document
corrects those errors.
DATES: Effective September 4, 1997.
FOR FURTHER INFORMATION CONTACT: Peter J. Eldridge, 813-570-5305.
SUPPLEMENTARY INFORMATION: The shrimp fishery off the southern Atlantic
states is managed under the Fishery Management Plan for the Shrimp
Fishery of the South Atlantic Region (FMP). The FMP was prepared by the
South Atlantic Fishery Management
[[Page 46680]]
Council and is implemented under the authority of the Magnuson-Stevens
Fishery Conservation and Management Act by regulations at 50 CFR part
622.
Background and Need for Correction
Amendment 2 to the FMP, among other things, requires the use of
certified BRDs in all penaeid shrimp trawls in the exclusive economic
zone in the South Atlantic. The final rule to implement Amendment 2
contained as an appendix the Testing Protocol for BRD Certification.
The Statistical Approach section of that appendix contained errors,
primarily in the formulas for computing whether the BRD tested achieves
the minimum required reduction rate for weakfish and Spanish mackerel.
For clarity and ease of use, this correction restates the entire
corrected Statistical Approach section of the Testing Protocol for BRD
Certification.
Correction of Publication
In FR Doc. 97-9816 published on April 16, 1997 (62 FR 18536), make
the following correction. On page 18542, in the appendix to the
document, which will not appear in the Code of Federal Regulations,
remove the text beginning five lines from the bottom of the first
column with the heading ``Statistical Approach'' and ending with the
text on line 11 of the third column and replace it with the following:
Statistical Approach
You should start with the assumption that the BRD to be tested
does not achieve the minimum required reduction rate, say
R0. This assumption will be rejected if the data provide
sufficient evidence to do so. Hence, the hypotheses to be tested are
as follows:
H0: BRD does not achieve the minimum required reduction
rate,
[GRAPHIC] [TIFF OMITTED] TR04SE97.008
Ha: BRD does achieve the minimum required reduction rate,
[GRAPHIC] [TIFF OMITTED] TR04SE97.009
Here R denotes the actual reduction rate (unknown),
R0 denotes the minimum required reduction rate,
denotes the actual mean CPUE with the
control, and denotes the actual mean
CPUE with the BRD.
With any hypothesis testing, there are two risks involved, known
as type I error (rejecting the true H0) and type II error
(accepting a false H0). The probabilities of committing
these errors are denoted by alpha and beta, respectively, and those
are inversely related to each other. As alpha increases, beta
decreases, and vice versa. The above test will be conducted with an
alpha to be specified by the RA. The above hypotheses should be
tested using a ``modified'' paired t-test.
The CPUE values for the control and BRD nets for each successful
tow should be computed first and these will be used in the following
computations. The test statistic to be used is given by:
[GRAPHIC] [TIFF OMITTED] TR04SE97.010
Where:
x is the observed mean CPUE for the control,
y is the observed mean CPUE for the BRD,
sd0 is the standard deviation of
di={(1-R0)xi-yi} values, n is the number of
successful tows used in the analysis, and i=1, 2, . . . , n.
The H0 will be rejected if t>talpha,n-1,
where talpha,n-1 denotes the (1-alpha) 100th percentile
score in the t distribution with (n-1) degrees of freedom.
The computation of beta (for various assumed reduction rates,
R1>R0) is somewhat involved and requires the
knowledge of unknown parameters (or at least good estimates) of
c and
20. Note that
20 is dependent on the
R0 specified (under H0) and equals:
(1-R0)2
2+2
-2(1-R0)
,
where is the population correlation coefficient between
xi and yi values. The computation of beta in
advance (in the absence of any preliminary data, i.e., without good
parameter estimates) is almost impossible. More work in this
direction is still needed. However, it is clear that beta could be
reduced by increasing alpha or n or both.
A (1-alpha) 100% two-sided confidence interval on R consists of
all values of R0 for which
H0: R=R0 (versus Ha:
RR0) cannot be rejected at the level of
significance of alpha. One-sided confidence intervals on R could
also be computed appropriately.
Dated: August 27, 1997.
David L. Evans,
Deputy Assistant Administrator for Fisheries, National Marine Fisheries
Service.
[FR Doc. 97-23404 Filed 9-3-97; 8:45 am]
BILLING CODE 3510-22-P
