Yu-ui
Overview
Yu-ui (有意) in statistics refers to a state in which an observed difference or relationship is judged unlikely to have occurred by chance, and it is determined through the significance level and the p-value. Meanwhile, Yu-ui written with the Chinese characters 留意 is also widely used in the everyday sense of "paying careful attention," and frequently appears in the form of "precautions" in laws, contracts, and pharmaceutical instructions. Although the two meanings are similar in pronunciation and spelling, in academic contexts they mostly refer to statistical significance, while in practical contexts they indicate attention and caution.
Main content
Etymology and linguistic meaning
The meaning of Yu-ui differs according to its Chinese character spelling. 留意 (Yu-ui) is a combination of "留" (to remain) and "意" (intention/meaning), meaning to hold one's mind on something and pay attention. Expressions such as "precautions," "points to note," and "please pay particular attention" fall into this category, and are used in public institution notices, product manuals, and legal documents as devices to limit the scope of responsibility or to remind users of their duty of care. 有意 (Yu-ui) is a combination of "有" (to have) and "意" (intention/meaning), meaning "to have meaning" or "to be meaningful," and together with the derived term "有意味" (meaningful), it has become a key term in academic and statistical contexts. In addition, 流議 means floating discussion or public commentary, a rare usage that is almost never used in modern Korean.
Significance in statistics
Statistical hypothesis testing is a procedure that calculates the probability that a result observed in a sample would appear from random variation alone. A researcher first establishes a null hypothesis of "no difference" and an alternative hypothesis of "there is a difference," calculates a test statistic, and then produces a p-value. If the p-value is smaller than a pre-set significance level α, the null hypothesis is rejected, and the result is said to be "statistically significant." By convention, α = 0.05 (5%) is most often used, and in pharmaceutical clinical trials requiring strict criteria, 0.01 or 0.025 (one-sided or two-sided tests) may be applied.
There are two types of errors in determining significance. A Type I error is when a significant result is incorrectly declared although there is actually no difference, and its probability is exactly the significance level α. A Type II error is when a real difference is not detected, and the value obtained by subtracting the Type II error probability from 1 is called power. Because power is determined by sample size, effect size, and significance level, calculating the sample size in advance at the study design stage has become a standard procedure.
Significance level and confidence intervals
The concept corresponding to the significance level α is the confidence interval. A 95% confidence interval has the frequentist meaning that 95% of intervals calculated by repeated sampling in the same way would contain the true value. If the confidence interval does not include the value meaning "no effect" (e.g., a difference of 0, a correlation coefficient of 0, or a risk ratio of 1), it is statistically significant at the 5% significance level. Therefore, in recent years it has been recommended to report effect sizes and confidence intervals together rather than presenting only a p-value.
Yu-ui in practice and administration
In administrative practice, "Yu-ui" functions as an expression specifying a duty of care. The precautions clause in a contract prevents disputes by notifying the other party of conditions that are easy to overlook, and the precautions in a package insert for a drug provide guidance on contraindications, side effects, and dosage. In safety guidelines at construction and manufacturing sites, expressions such as "objects requiring caution" and "matters requiring attention" also appear repeatedly, and these go beyond simple recommendations to sometimes serve as grounds for judging legal responsibility.
Yu-ui in medicine and clinical practice
In clinical research, there is active discussion that "statistical significance" and "clinical significance" must be distinguished. This is because if the sample is sufficiently large, even a tiny difference that patients do not actually feel can come out as statistically significant. To supplement this, the concept of the minimal clinically important difference (MCID) is used, and indicators that patients directly experience, such as quality of life, survival rate, and readmission rate, are evaluated together.
Misunderstandings and limitations
A p-value is neither the probability that the null hypothesis is true nor the probability that the result is due to chance. In addition, statistical significance does not automatically guarantee the importance or practical usefulness of a result. If multiple hypotheses are repeatedly tested without correcting for multiple comparisons, significant results can appear by chance alone, and publication bias—hiding non-significant results and reporting only significant ones—undermines the credibility of the literature as a whole.
Recent trends
In 2016, the American Statistical Association (ASA) issued a statement advising against using p-values as a dichotomous criterion, and as discussions of the replication crisis have since spread, research practices have been changing rapidly. As of 2024–2025, pre-registration, disclosure of analysis plans, and sharing of data and code have become standard requirements in major journals, and reporting centered on effect sizes and confidence intervals has become common. Bayesian approaches, equivalence testing, and multiple-comparison correction based on FDR (false discovery rate) are also widely used in life science and omics research. In addition, as machine learning and large-scale data analysis have become widespread, awareness has grown that "statistical significance" alone cannot guarantee explanatory power, and the practice of presenting predictive performance, cross-validation, and external validation together has taken root. In Korea as well, clinical trial statistical guidelines and research ethics regulations have been strengthened, and there is a trend requiring explicit reporting of the rationale for setting the significance level and the sample size calculation process.
Related topics
- [[Statistics]]
- [[Hypothesis testing]]
- [[p-value]]
- [[Significance level]]
- [[Replication crisis]]
- [[Confidence interval]]
- [[Error]]
- [[Duty of care]]