Yuen's Trimmed Means t Test

Yuen's trimmed means t test compares two independent groups using trimmed means and winsorized variances, providing a robust alternative to the Welch t test under outliers and skewness.

Yuen 절사평균 t 검정은 절사평균과 윈저화 분산으로 두 독립 집단을 비교해, 이상치와 치우침에 견고한 Welch t 검정 대안을 제공한다.

My two groups have outliers and skewed scores; is there a robust version of the t test?

두 집단에 이상치가 있고 점수가 한쪽으로 치우쳐 있는데, 강건한 t 검정이 있나요?

Yuen's test removes a fixed proportion, usually 20 percent, from each tail of each group and compares the remaining means. Its standard error uses winsorized variances and its degrees of freedom follow a Welch-type approximation, so equal variances are not assumed.

When to use Yuen's Trimmed Means t Test

• Two independent groups are compared on a continuous outcome with outliers.
• Distributions are skewed or heavy-tailed.
• A robust location comparison is preferred to a rank test.

When to use something else

• Data are close to normal without outliers; the Welch t test is more efficient.
• The two measurements come from the same participants; use a paired Yuen test or the Wilcoxon signed-rank test.

Assumptions and requirements

• Observations are independent within and between groups.
• Each group retains at least two observations after trimming.
• Trimmed population means are the parameters of interest.

What the analysis reports

• t statistic
• Welch-Yuen degrees of freedom
• p value
• Trimmed mean difference with confidence interval
• Group trimmed means and effective sizes

How to report it

Yuen's test on 20% trimmed means indicated that group 1 (M_t = [tm1]) and group 2 (M_t = [tm2]) differed significantly, t([df]) = [t], p = [p], 95% CI [[lower], [upper]].

Related methods

• Welch's t-test
• Mann-Whitney U Test
• Robust ANOVA on Trimmed Means

Run Yuen's Trimmed Means t Test on Quorant