Đang chuẩn bị liên kết để tải về tài liệu:
Analysis of Survey Data phần 4

Như đã nói từ hình 10.1, chỉ số BMI và DBMI có thể có liên quan. Điều này có thể được theo đuổi bằng cách kiểm tra dự toán của hàm mật độ doanh. Hình 10.4 cho thấy một âm mưu 3-D và vẽ một đường viền cho f (x, y) cho phụ nữ, trong đó x là BMI và y là DBMI. | 100 CATEGORICAL RESPONSE DATA FROM COMPLEX SURVEYS Table 7.2 Rankings of test performance based on Type I error and power. Average ESLs Average EPs 0.3 a 0.6 4.8 ESL 7.0 0.7 a 1.0 5.4 ESL 7.0 55 EP 73 FX 5. a FXW LL EV 1 k 0.5 F X 5. Bf LL FXjW LL EV 1 k 0.5 F X 5. Bf LL FQ T LL 0.05 X 5. a F 22 à. a fQ T LL 0.05 X 5. a fXW LL F X 5. X 5. XLRJa Bf LL FQ T LL 0.05 fXW LL Xlrj FXW- LL EV 1 k 0.5 FX 5. F X 5. FX 5. a a For L 30 to avoid error rate inflation for small L. and second Servy Hachuel and Wojdyla 1998 included the Morell modified Wald procedure XW LL M in their list. They gave the second-order Rao Scott statistic Xp S. a a higher power ranking than did Thomas Singh and Roberts 1996 but an important point on which both studies agreed is that the log linear Bonferroni procedure Bf LL is the most powerful procedure overall. Thomas Singh and Roberts 1996 also noted that Bf LL provides the highest power and most consistent control of Type I error over tables of varying size 3 X 3 3 X 4 and 4 X 4 . 7.3.3. Discussion and final recommendations The benefits of Bonferroni procedures in the analysis of categorical data from complex surveys were noted by Thomas 1989 who showed that Bonferroni simultaneous confidence intervals for population proportions coupled with log or logit transformations provide better coverage properties than competing procedures. This is consistent with the preeminence of Bf LL for tests of independence. Nevertheless it is important to bear in mind the caveat that the log linear Bonferroni procedure is not invariant to the choice of basis for the interaction terms in the log linear model. The runner-up to Bf LL is the Singh procedure FQT LL which was highly rated in both studies with respect to power and Type I error control. However some practitioners might be reluctant to use this procedure because of the difficulty of selecting the value of e. For example Thomas Singh and Roberts 1996 recommend e 0.05 while Servy Hachuel and Wojdyla 1998 .