TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

On the hazards of significance testing: the false discovery rate

25 点作者 jonathansizz大约 10 年前

1 comment

sukilot大约 10 年前
In short, p=0.05 means that 5% of no-effect experiments show a false positive.<p>But due to publication bias, negatives are ignored. When there are lots of experiments looking for mostly non-existent effects , the false positives appear ever more commonly than true positives.<p>This is why Bayesians insist on choosing a prior probability. Without it, there is no way to interpret the result of an experiment.<p>The takeaway lesson here, which is horribly underemphasized in school , is that statisitics never answers yes&#x2F;no questions! Statistics can only give you a <i>function</i> for combining a Prior with an Experiment to obtain a Posterior. Without a Prior (which is subjective; it can never be chosen objectively!), you cannot get a Posterior Probability out of a experiment.