Years on the NLS team: 1995-present
Role: As the sampling statistician for NLSY97, I selected the sample of over 90,000 housing units that were visited to collect the base year interviews. I also created the weights and the weighting program engine still used to calculate the weights.
The sample design for NLSY97 was more efficient in collecting minority youth interviews than in 1979 through the simple idea of using counties as the first-stage sample units for the 1997 Supplemental sample. This idea was expanded for another project described in Pedlow (2014, Journal of Official Statistics). During Round 4, Colm O'Muircheartaigh and I improved the weights by switching from simply combining the Cross-Sectional and Supplemental samples to cumulating the selection probabilities to unify the samples, which decreased standard errors for NLSY97 minority estimates by 20 percent (O'Muircheartaigh and Pedlow, 2002 Proceedings of the American Statistical Association Survey Research Methods Section). NLSY97 was also one of two major NORC projects that helped us build a linear programming method to efficiently assign cases to field interviewers (English and Pedlow, 2007 Proceedings of the American Statistical Association Survey Research Methods Section).
The NLSY97 was the first major project of my career. Working with Martin R. Frankel and Kirk Wolter on the sample design taught me a lot, and put me on the path I am still on at NORC. I feel very fortunate to be a part of one of the world's richest datasets, and I have greatly enjoyed working on this project through the years.