Cohen d-ning Effekt Hajmini Talqin Qilish

Interaktiv vizualizatsiya

Tomonidan yaratilgan Kristoffer Magnusson · Version 2.7.0 ·

Cohen d-ning effekt hajmi psixologiyada juda mashhur. Biroq, uning talqini oddiy emas va tadqiqotchilar ko'pincha effektni sharhlashda kichik (0,2), o'rta (0,5) va katta (0,8) kabi umumiy ko'rsatmalardan foydalanadilar. Bundan tashqari, ko'p hollarda standartlashtirilgan o'rtacha farq standartlashtirilmagan o'rtacha farqdan ko'ra ko'proq izohlanadimi, degan savol tug'iladi.

Cohen d-ni talqin qilishga yordam berish uchun ushbu vizualizatsiya Cohen d-ning turli xil tasvirlarini taqdim etadi: vizual qoplama, Cohen U₃, ustunlik ehtimoli, bir-biriga mos kelish foizi va raqam davolash uchun zarur. Shuningdek, u standart og'ishni o'zgartirishga imkon beradi va standartlashtirilmagan farqni ko'rsatadi.

Cohen d-ning

Drag the chart to pan · double-click to recenter.

Cohen U₃-ning

% Qoplash

Ustunlik ehtimoli

Davolash uchun zarur bo'lgan raqam

Umumiy til izohi

Cohen d 0.80 ga teng bo'lsa, "Davolash" guruhining 78.8% "Nazorat" guruhining o'rtacha qiymatidan yuqori bo'ladi (Cohen U₃), 68.9% ikkita guruh bir-biriga mos keladi va davolash guruhidan tasodifiy tanlangan kishi nazorat guruhidan tasodifiy tanlangan kishiga qaraganda yuqori ballga ega bo'lish ehtimoli 71.4% (ustunlik ehtimoli). Bundan tashqari, davolash guruhida nazorat guruhiga nisbatan yana bitta ijobiy natijaga erishish uchun biz o'rtacha 3.5 kishini davolashimiz kerak. Bu shuni anglatadiki, agar har bir guruhda 100 kishi bo'lsa va biz nazorat guruhida 20 kishi ijobiy natijaga ega deb hisoblasak, davolash guruhidagi 20 + 28.3 kishi ijobiy natijalarga erishadi.¹

¹The values are averages, and it is assumed that 20 (CER) of the "Nazorat" group have "favorable outcomes," i.e., their outcomes are below some cut-off. Change this using the Settings button above the slider. Go to the formula section for more information.

FAQ

Ushbu vizualizatsiyadan qanday foydalanaman?

Cohen d-ni o'zgartirish uchun slayderdan foydalaning yoki sozlamalar panelini oching va parametrlarni o'zgartiring. Kirishlarni klaviatura strelkalari yordamida ham boshqarish mumkin.

From the settings you can change the means, standard deviation, and the control group event rate (CER); the axis and distribution labels; and the slider's maximum and step size. Settings can be saved in your browser's localStorage so they persist across visits.

Displeyni bosish va uni sudrab borish orqali X o'qini panorama qilishingiz mumkin. Vizualizatsiyani markazlashtirish va hajmini o'zgartirish uchun ikki marta bosing. Displayni bosish va sudrab borish orqali x o'qini panorama qilishingiz mumkin. Vizualizatsiyani markazlashtirish va o‘lchamini o‘zgartirish uchun ikki marta bosing.

Formulalar qanday?

Cohen's d is the standardized mean difference, where $\delta$ is the population Cohen's d, equal population variances ( $\sigma$ ) are assumed, and $\mu_i$ is each population mean:

\delta = \frac{\mu_2 - \mu_1}{\sigma}

Writing $\Phi$ for the standard-normal CDF and $\Psi$ for its inverse, Cohen's U₃ (Cohen, 1977), the overlapping coefficient (Reiser & Faraggi, 1999), the probability of superiority / common-language effect size (Ruscio, 2008), and the number needed to treat (Furukawa & Leucht, 2011) are:

U_3 = \Phi(\delta)

\text{OVL} = 2\,\Phi(-|\delta|/2)

\text{CL} = \Phi\!\left(\frac{\delta}{\sqrt{2}}\right)

\text{NNT} = \frac{1}{\Phi(\delta + \Psi(\text{CER})) - \text{CER}}

CER is the control group's event rate (20% by default; change it in the settings). The overlap statistic here intentionally differs from Cohen's original calculation — see the question below.

R code to calculate NNT from Cohen's d:

CER <- 0.2
d <- 0.2
1 / (pnorm(d + qnorm(CER)) - CER)

Iqtiboslar

Cohen, J. (1977). Statistical power analysis for the behavioral sciences. Routledge.
Furukawa, T. A., & Leucht, S. (2011). How to obtain NNT from Cohen's d: comparison of two methods. PLoS ONE, 6(4), e19070.
Reiser, B., & Faraggi, D. (1999). Confidence intervals for the overlapping coefficient: the normal equal variance case. Journal of the Royal Statistical Society, 48(3), 413–418.
Ruscio, J. (2008). A probability-based measure of effect size: robustness to base rates and other factors. Psychological Methods, 13(1), 19–30.

Ushbu sahifadan qanday qilib iqtibos keltiraman?

Cite this page according to your favorite style guide.

The current version is 2.7.0 (changelog).

APA 7

Magnusson, K. (2026). Interpreting Cohen's d effect size: An interactive visualization (Version 2.7.0) [Web App]. R Psychologist. https://rpsychologist.com/cohend/

BibTeX

@software{magnussonCohend,
  author = {Kristoffer Magnusson},
  title = {Interpreting Cohen's d effect size: An interactive visualization},
  url = {https://rpsychologist.com/cohend/},
  version = {2.7.0},
  date = {2026-10-24},
}

Men bug/xato/matbaa xatosini moliyalashtiraman yoki taklif qilmoqchiman!

Please open an issue on GitHub, or reach out on social media.

Bir-biriga o'xshashlik statistikasi Cohen-ning hisob-kitoblaridan farq qiladi

Bu ataylab qilingan, mening sabablarim haqida ushbu blog postda ko'proq o'qishingiz mumkin: Where Cohen went wrong – the proportion of overlap between two normal distributions

Ushbu vizualizatsiyani kitobim/maqolam/va hokazolarimga qo'shishim mumkinmi?

Ha, davom eting! Men ikkita bir-biriga o'xshash Gauss taqsimotini tuzishni o'ylab topmaganman. Bu vizualizatsiya ommaviy domenga bag'ishlangan bo'lib, bu "siz ishni ko'chirish, o'zgartirish, tarqatish va hatto tijorat maqsadlarida ham ruxsat so'ramasdan bajarishingiz mumkin" degan ma'noni anglatadi (Creative common CC0-litsenziya-siga qarang). Garchi atribut talab qilinmasa ham, u har doim qadrlanadi! Ushbu sahifaning manba kodi MIT yordamida litsenziyalangan va sahifadagi matn CC-BY 4.0.

Ko'proq vizualizatsiya

All visualizations →