:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

:^(

想問meta analysis唔入non significant findings 去分析會有咩bias? 因為其實入哂non significant findings最後個combined effect可能變返significant result. 咁有咩意思?

呢個問題存在好耐
open database 係解決之道

呢個情況其實似你普通run individual data時
你放兩組15+15人時唔sig, 放30+30時可能sig, 每個人自己baseline posttest比可能唔食

一切係effect size and power既問題

如果你唔放non sig data, 一來係publication bias selection bias,
二來驗證唔到個effect size係因為power定其他

當然，如果你越放得多越heterogeneous, 咁代表不同paper variety 高

講到尾，meta analysis都只不過係一種estimation，我唔明點解咁多人覺得佢好神

:^(

sor, 有感而發

:^(

:^(