Scholarlink

<p>Statistical significance is often treated as evidence of importance in quantitative business research, but this inference is problematic in large observational datasets, where even trivial effects can be estimated precisely. We examine this problem in a social media context, where prior studies report statistically significant effects of emotional, linguistic, and multimedia features of a content on the users’ sharing behavior. Using 9,326 Mashable news articles linked through the outlet’s official Facebook posts, we assess whether these effects are substantively meaningful and whether they generalize across editorial sections. The findings point to informative substantive nulls. Although several content-related predictors are statistically significant, together they explain only 1.6% of the variance in sharing behavior. Additionally, individual effects vary in both magnitude and direction across content domains, limiting their contextual generalizability. These results suggest that the literature may overstate the importance of content-related drivers of sharing when statistical detectability is conflated with substantive contribution. The study contributes to ongoing discussions about substantive significance by showing how partial <i>R</i>² decomposition and outcome-based effect translation can help researchers distinguish statistically significant predictors from predictors that are substantively meaningful in social media and marketing analytics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

When significance misleads: content factors explain little of content popularity

  • 期刊论文

  • 发表日期：

    2026年7月30日

查看全文
  • Tohid Ghanbarpour,
  • Mona Solvoll

摘要

Statistical significance is often treated as evidence of importance in quantitative business research, but this inference is problematic in large observational datasets, where even trivial effects can be estimated precisely. We examine this problem in a social media context, where prior studies report statistically significant effects of emotional, linguistic, and multimedia features of a content on the users’ sharing behavior. Using 9,326 Mashable news articles linked through the outlet’s official Facebook posts, we assess whether these effects are substantively meaningful and whether they generalize across editorial sections. The findings point to informative substantive nulls. Although several content-related predictors are statistically significant, together they explain only 1.6% of the variance in sharing behavior. Additionally, individual effects vary in both magnitude and direction across content domains, limiting their contextual generalizability. These results suggest that the literature may overstate the importance of content-related drivers of sharing when statistical detectability is conflated with substantive contribution. The study contributes to ongoing discussions about substantive significance by showing how partial R² decomposition and outcome-based effect translation can help researchers distinguish statistically significant predictors from predictors that are substantively meaningful in social media and marketing analytics.