Jonckheere-Terpstra tests for doubly ordered R×C tables were used. If 20 percent of cells had expected counts of less than 5, exact p values were calculated by means of data permutation. If the calculation took longer than the StatXact software’s default limit of 30 seconds, Monte Carlo sampling (10,000 samples) was used to estimate the exact p values.
For this study, alpha was set at 0.05. A Bonferroni corrected alpha value to adjust for the family-wise error rate of performing the 15 statistical tests of this study results in an alpha of 0.003. The use of the Bonferroni adjusted alpha value decreases the potential for Type I errors, but increases the potential for Type II errors. Unadjusted p values are reported in this article. Statistical analyses were performed with JMP Statistical Software (Release 126.96.36.199, SAS Institute, Inc., Cary, NC) and StatXact Statistical Software for Exact Nonparametric Inference (Version 6, Cytel Software Corp., Cambridge, MA).
Ethics approval was obtained from the Washington University Medical School Human Studies Committee.
The authors received surveys from 575 practices, which represent 1,006 radiologists or 64 percent of 1,572 actively practicing breast imagers in the SBI. This return rate provided a 99 percent level of confidence for report responses, with a confidence interval of ±3 percentage points (Rea and Parker, 1997). The sample included practices with a broad range of case volumes and serving a diverse population. The number of responses varied with each item.
Farria D, Schmidt ME, Monsees BS, Smith RA, Hildebolt C, Yoffie R, Monticciolo DL, Feig SA, Bassett LW. In press. Professional and economic factors affecting access to mammography: A crisis today, or tomorrow? Results from a national survey. Cancer.
Rea LM, Parker RA. 1997. Designing and Conducting Survey Research: A Comprehensive Guide. 2nd ed. San Francisco, CA: Jossey-Bass. Pp. 118–123.