演講公告
首頁 \ 學術交流 \ 演講公告
講座名稱 Evaluating the replicability of psychological science: A meta-analytic moderator analysis approach
主講人 顏志龍 
講座日期 2017-10-27
講座時間 13:10-15:00
講座地點 北棟2樓階梯教室
英文演說
主辦系所 心理學系
備註  演講相關內容如下:

=====================================================================

時間:10/27()下午13:10-15:00

講者:顏志龍老師 (銘傳大學 諮商與工商心理學系 教授 

地點:社科大樓北棟2樓心理學系階梯教室

個人簡介:

-學歷:國立政治大學心理學系博士

-學術專長:實驗心理學、研究方法、社會心理學

-相關著作:

 

http://www.chihlongyen.url.tw/About/yan-jiu-ji-zhu-zuo/

 

 

-個人網站:http://www.chihlongyen.url.tw/

講題:Evaluating the replicability of psychological science: A meta-analytic moderator analysis approach

摘要:

Recently, the validity of some psychology findings has been questioned. In evaluating the replicability of research findings, the approach typically used has been to directly replicate selected studies. However, this approach of direct replication via individual studies has been criticized for having insufficient statistical power, for failing to take sampling variability into account, and for requiring excessive resource consumption. In this study, we evaluated replicability through re-analyses of various meta-analytic studies in the past literature. Considering the nature of replication, the replicated effect size should be largely affected by the  research practices of the replication study (e.g., measurements or manipulation of independent/dependent variables, sample characteristics, etc.). More specifically, the more similar a replication study and the study it seeks to replicate are in terms of their research practices, the higher the probability that the results can be replicated. Therefore, if studies are replicable, variation in research practices should account for the variation in effect sizes. We screened twenty years (1997-2016) of meta-analysis articles published iPsychological Bulletin and reanalyzed 90 meta-analyses using meta-analytic moderation analysis. The results indicated that, on average, variation in research practices can explain less than half of effect size variance (R2 = .49). Under such circumstances, there is a 42% chance that the discrepancy in effect sizes between the original and the replication studies will be larger than a small effect. These results implied that replication might still fail even if a replication study closely follows the research practices of the study it seeks to replicate.

=====================================================================