(a) Use test results from two engines for each engine family to calculate the required sample size for the model year. Update this calculation with each test.

(b) Early in each calendar quarter, randomly select and test two engines from the end of the assembly line for each engine family.

(c) Calculate the required sample size for each engine family. Separately calculate this figure for HC+NOX and for CO. The required sample size is the greater of these two calculated values. Use the following equation: $$

Where:

N = Required sample size for the model year.

t95 = 95% confidence coefficient, which depends on the number of tests completed, n, as specified in the table in paragraph (c)(1) of this section. It defines 95% confidence intervals for a one-tail distribution.

x = Mean of emission test results of the sample.

STD = Emission standard.

σ = Test sample standard deviation (see paragraph (c)(2) of this section).

(1) Determine the 95% confidence coefficient, t95, from the following table:

------------------------------------------------------------------------
    n t[INF]95[/INF]         n t[INF]95[/INF]        n t[INF]95[/INF]
------------------------------------------------------------------------
 2 6.31                  12 1.80                  22 1.72
 3 2.92                  13 1.78                  23 1.72
 4 2.35                  14 1.77                  24 1.71
 5 2.13                  15 1.76                  25 1.71
 6 2.02                  16 1.75                  26 1.71
 7 1.94                  17 1.75                  27 1.71
 8 1.90                  18 1.74                  28 1.70
 9 1.86                  19 1.73                  29 1.70
10 1.83                  20 1.73                  30+ 1.70
11 1.81                  21 1.72                  ......................
------------------------------------------------------------------------

(2) Calculate the standard deviation, σ, for the test sample using the following formula: $$

Where:

Xi = Emission test result for an individual engine.

n = The number of tests completed in an engine family.

(d) Use final deteriorated test results to calculate the variables in the equations in paragraph (c) of this section (see §1048.315(a)).

(e) After each new test, recalculate the required sample size using the updated mean values, standard deviations, and the appropriate 95-percent confidence coefficient.

(f) Distribute the remaining engine tests evenly throughout the rest of the year. You may need to adjust your schedule for selecting engines if the required sample size changes. Continue to randomly select engines from each engine family; this may involve testing engines that operate on different fuels.

(g) Continue testing any engine family for which the sample mean, x, is greater than the emission standard. This applies if the sample mean for either HC+NOX or for CO is greater than the emission standard. Continue testing until one of the following things happens:

(1) The sample size, n, for an engine family is greater than the required sample size, N, and the sample mean, x, is less than or equal to the emission standard. For example, if N = 3.1 after the third test, the sample-size calculation does not allow you to stop testing.

(2) The engine family does not comply according to §1048.325.

(3) You test 30 engines from the engine family.

(4) You test one percent of your projected annual U.S.-directed production volume for the engine family.

(5) You choose to declare that the engine family does not comply with the requirements of this subpart.

(h) If the sample-size calculation allows you to stop testing for a pollutant, you must continue measuring emission levels of that pollutant for any additional tests required under this section. However, you need not continue making the calculations specified in this section for that pollutant. This paragraph (h) does not affect the requirements in section §1048.320.

(i) You may elect to test more randomly chosen engines than we require. Include these engines in the sample-size calculations.