
An official website of the United States government
This paper highlights two issues that came up in recent work on a large set of employment series from the Current Employment Statistics program in the Bureau of Labor Statistics. The first examines performing hypotheses tests on such a large set of series. As seasonal adjustment practitioners move from empirical to model-based diagnostics, care needs to be taken to mitigate the effects of multiple hypothesis testing on this set of 4700+ series. The second topic involves using the model-based seasonal F-test to determine if a series is seasonal. Since models used in seasonal adjustment production rarely include seasonal regressors, a second X-13 run is often needed to generate this test. This work examines two methods for generating the model for this second run, and what difference this makes to the results of the seasonal F-test.