10 C.F.R. § 429.25
(a) Sampling plan for selection of units for testing.
(2) For each basic model of television, samples shall be randomly selected and tested to ensure that—
(i) Any represented value of power consumption of a basic model for which consumers would favor lower values shall be greater than or equal to the higher of:
(A) The mean of the sample, where:

and x is the sample mean; n is the number of samples; and xi is the i th sample;
Or,
(B) For on mode power consumption, the upper 95 percent confidence limit (UCL) of the true mean divided by 1.05, where:

and x is the sample mean; s is the sample standard deviation; n is the number of samples; and t0.95 is the t-statistic for a 95% one-tailed confidence interval with n−1 degrees of freedom (from appendix A of this subpart).
And
(C) For standby mode power consumption and power consumption measurements in modes other than on mode, the upper 90 percent confidence limit (UCL) of the true mean divided by 1.10, where:

and x is the sample mean; s is the sample standard deviation; n is the number of samples; and t0.90 is the t-statistic for a 90% one-tailed confidence interval with n−1 degrees of freedom (from appendix A of this subpart).
(iii) Rounding requirements. The represented value of power consumption and the represented annual energy consumption shall be rounded as follows:
[78 FR 63840, Oct. 25, 2013, as amended at 88 FR 16109, Mar. 15, 2023]