1 Weighting Difficulties with a Sample from a National Survey Background A cross-sectional, retrospective, explanatory secondary analysis of the 2002 National Health Interview Survey was conducted in order to describe complementary and alternative modality (CAM) use and high-risk CAM use by women with female-specific cancers. Methodologic Challenge Initially, the researcher planned to weight the sample for estimations in order to be able to generalize the findings to the population of women in the US with female-specific cancers. However, when running data on Stata with PSU, strata, and person weights, an error message was given, stating that there were strata with single sampling units. When this is the case, weighted estimations cannot be calculated. The researcher then proceeded to delete cases from the sample of 598 women in which there was a single PSU per stratum. This yielded a sample of only 163 women. Weighted estimations from such a small initial sample size, which was no longer representative of those in the US with female-specific cancers, was considered by the researcher to be unacceptable for estimations. The researcher considered using the person weights alone, as a solution to bypassing the problem of strata with single PSUs. In advice given by Dr. Jeffrey Rhoades from the Agency for Healthcare Research and Quality, “ignoring strata and/or PSU is really, really bad. Even though you can get correct point estimates, the test, confidence intervals, and standard errors will usually be wrong.” Rhoades indicated that Stata software has less options for dealing with this situation than SUDAAN or R survey software; thus, the best solution was to avoid single PSUs. He stated the recommendation from Stata documentation is to collapse strata, but noted that collapsing strata is “extremely dangerous when you don’t know how the strata are formed.” Thus, the researcher used unweighted data for estimations of association and prediction. Valerie Eschiti, PhD, RN, CHTP, AHN-BC is an assistant professor at the College of Nursing, University of Oklahoma Health Sciences Center, Oklahoma City, OK. She can be contacted at valerie-eschiti@ouhsc.edu.