愛華網(wǎng)
Meta分析時(shí)，雖然數(shù)據(jù)類型有各種，算法不同，其實(shí)歸根結(jié)底都要得到效應(yīng)量及其標(biāo)準(zhǔn)誤，如果只有效應(yīng)量和可信區(qū)間，可以根據(jù)效應(yīng)量及可信區(qū)間求出標(biāo)每個(gè)研究的準(zhǔn)誤，如以RR為例，其對(duì)數(shù)標(biāo)準(zhǔn)誤為se=(loguci-logRR)/1.96，UCI是區(qū)間中的最高值，獲得每個(gè)研究的RR及其標(biāo)準(zhǔn)誤，就比較容易進(jìn)行Meta分析了。
如RR假設(shè)其可信區(qū)間為95%，則stata 命令如下：
gen selogrr=(log(uci)-log(rr))/1.96
gen logrr=log(rr)
metan logrr selogrr,eform by(group) label(namevar=study)
一篇英文文獻(xiàn)這樣寫：If 95 percent confidence limits are given instead ofstandard errors and the limits are arithmetically symmetric aboutthe rates, i.e., if R - LR = UR - R, the standard errors may beestimated as SE =(UR - RL)/3.92, where LR and UR are the upperandlower confidence limits. If 95 percent confidence limits are givenand the limits appear proportionally symmetric about the rates,i.e., if U/R = R/LR, the standard errors for the_ rates may beestimated as SE =R[log(UR/LR)/3.92].
