Abstract：
OBJECTIVE: Reliance on null hypothesis significance testing often leads to misinterpretation of research results. Common misinterpretations include that a statistically nonsignificant difference (p≥0.05) implies no difference between groups, and that a statistically significant finding (p<0.05) is unbiased and clinically important. We aimed to develop a tool - the Conclusion Generator - to mitigate these misconceptions.
METHODS: We reviewed the content of the Conclusion Generator and validated its output using published and simulated data.
RESULTS: The Conclusion Generator is a free online application designed to generate conclusions for scientific papers based on the values and clinical interpretation of the point estimate and confidence interval. Both relative and absolute measures of effect are supported. It offers two modes for interpretation: (1) Statistical mode provides an accurate statistical interpretation of results, with an optional specification of superiority and noninferiority bounds; (2) Clinical mode evaluates the clinical importance of the point estimate and confidence limits as specified by the user. Both modes assume no uncontrolled biases. Users must specify the number of decimals, the direction of a beneficial effect (e.g., relative risk <1 vs. >1), and the level of detail (concise vs. elaborated) for the output. The validation confirmed the Conclusion Generator\'s capability to interpret research results, considering random error and clinical relevance, while avoiding common misinterpretations associated with null hypothesis significance testing.
CONCLUSIONS: The Conclusion Generator facilitates an appropriate interpretation of research results by emphasizing estimation and clinical relevance over hypothesis testing.
METHODS: We reviewed the content of the Conclusion Generator and validated its output using published and simulated data.
RESULTS: The Conclusion Generator is a free online application designed to generate conclusions for scientific papers based on the values and clinical interpretation of the point estimate and confidence interval. Both relative and absolute measures of effect are supported. It offers two modes for interpretation: (1) Statistical mode provides an accurate statistical interpretation of results, with an optional specification of superiority and noninferiority bounds; (2) Clinical mode evaluates the clinical importance of the point estimate and confidence limits as specified by the user. Both modes assume no uncontrolled biases. Users must specify the number of decimals, the direction of a beneficial effect (e.g., relative risk <1 vs. >1), and the level of detail (concise vs. elaborated) for the output. The validation confirmed the Conclusion Generator\'s capability to interpret research results, considering random error and clinical relevance, while avoiding common misinterpretations associated with null hypothesis significance testing.
CONCLUSIONS: The Conclusion Generator facilitates an appropriate interpretation of research results by emphasizing estimation and clinical relevance over hypothesis testing.
摘要：
目的：对零假设显著性检验的依赖往往导致对研究结果的误解。常见的误解包括统计学上无显著差异(p≥0.05)意味着组间没有差异，并且具有统计学意义的发现（p<0.05）是无偏的并且具有临床意义。我们旨在开发一种工具-结论生成器-来减轻这些误解。
方法：我们回顾了结论生成器的内容，并使用已发布和模拟的数据验证了其输出。
结果：结论生成器是一个免费的在线应用程序，旨在根据点估计和置信区间的值和临床解释为科学论文生成结论。支持相对和绝对效果度量。它提供了两种解释模式:(1)统计模式提供了对结果的准确统计解释,具有优势和非劣效性界限的可选规范;(2)临床模式评估用户指定的点估计和置信限度的临床重要性。两种模式都没有不受控制的偏见。用户必须指定小数的数量，有益效果的方向(例如，相对风险<1vs.>1),和详细程度(简洁与阐述)用于输出。验证证实了结论生成器解释研究结果的能力，考虑到随机误差和临床相关性，同时避免与零假设显著性检验相关的常见误解。
结论：结论生成器通过强调估计和临床相关性而不是假设检验，促进了对研究结果的适当解释。
方法：我们回顾了结论生成器的内容，并使用已发布和模拟的数据验证了其输出。
结果：结论生成器是一个免费的在线应用程序，旨在根据点估计和置信区间的值和临床解释为科学论文生成结论。支持相对和绝对效果度量。它提供了两种解释模式:(1)统计模式提供了对结果的准确统计解释,具有优势和非劣效性界限的可选规范;(2)临床模式评估用户指定的点估计和置信限度的临床重要性。两种模式都没有不受控制的偏见。用户必须指定小数的数量，有益效果的方向(例如，相对风险<1vs.>1),和详细程度(简洁与阐述)用于输出。验证证实了结论生成器解释研究结果的能力，考虑到随机误差和临床相关性，同时避免与零假设显著性检验相关的常见误解。
结论：结论生成器通过强调估计和临床相关性而不是假设检验，促进了对研究结果的适当解释。
