Commerce Tells CIT Threshold in 'd' Test Need Not Observe Statistical Assumptions

The Commerce Department stuck with its use of the Cohen's d test as part of its effort to root out "masked" dumping in an antidumping review after adding certain academic literature to the record as instructed by the Court of International Trade. Submitting its remand results to the trade court June 27, Commerce said certain statistical assumptions -- normality of the distribution, equal variances and around the same sample size -- don't limit the agency's use of the d test, given that it used the entire population of data as opposed to a sample (Nexteel Co. v. U.S., CIT Consol. # 18-00083).

Commerce added that its selection of the 0.8 threshold originally was set by statistician Jacob Cohen, the test's founder. While exporter SeAH Steel claimed the selection of this threshold required the implication of the key statistical assumptions, the agency said that Cohen did not set the threshold based on research results. Instead, Cohen set the 0.8 threshold because he said others would find the number, while arbitrary, to be reasonable.

The agency said the thresholds have been "widely accepted" and provide a "good basis for interpreting effect size and for resolving disputes about the importance of one's results." The literature also "provides no evidence that the values themselves or their use are dependent on statistical analysis or the application of the statistical criteria as argued by SeAH," Commerce said. "Indeed, their usefulness is based on their acceptance within the academic community."

As commonly argued in all cases concerning the Cohen's d test, Commerce said the statistical assumptions need not be observed because the test in the present case was run on the entire population of data. The agency said the literature centers on the application of the text using samples "and does not contain any express mention of criteria or assumptions necessary when examining an entire population," adding that if researchers look at the whole population, "these assumptions become unnecessary, as there is no need to make a whole population more reflective of the population."

The trade court had remanded the case, which concerns the 2015-16 administrative review of the AD order on oil country tubular goods from South Korea, to include academic literature cited by SeAH (see 2304200048). Judge Jennifer Choe-Groves said use of the 0.8 threshold, above which Commerce found masked dumping, was not squared with questions raised by the U.S. Court of Appeals for the Federal Circuit over the fact that the agency did not satisfy certain statistical assumptions when using the d test.