TABLE 3.
Begg and Mazumdar’s | Sterne and | Excess | ||||||
Trim-and-fill | rank correlation | Egger regression | significance test | p-uniform | p-curve | Selection models | PET-PEESE | |
Crude differences (k = 29) | ||||||||
Declines | 1(7%) | 2(13%) | 4(27%) | 1(7%) | 1(7%) | 0(0%) | 4(27%) | 2(13%) |
Increases | 1(11%) | 1(11%) | 2(22%) | 0(0%) | 2(25%) | 0(0%) | 3(33%) | 3(33%) |
Proteus | 3(60%) | 0(0%) | 2(40%) | 0(0%) | 0(0%) | 0(0%) | 2(40%) | 2(40%) |
Regression coefficients (k = 24) | ||||||||
Declines | 1(7%) | 1(7%) | 2(13%) | 1(7%) | 2(13%) | 0(0%) | 5(33%) | 2(13%) |
Increases | 1(11%) | 2(22%) | 4(44%) | 0(0%) | 1(12%) | 0(0%) | 2(22%) | 3(33%) |
Cell entries indicate absolute (and relative) frequencies of bias indication; k = 29/24, excepting p-uniform (k = 27/23), and p-curve (k = 26/23). k Declines = 15, k Increases = 9, k Proteus = 5.