Cohen's h

[Robert Murphy]

How can I compute Cohen's h to show effect size of intervention for two proportions?

[Bruce Weaver]

Hello Robert. Here's some code that duplicates the examples shown on this web-page:

. * https://people.ucalgary.ca/~ramsay/cohen-effect-size-h-arcsin-transformation.htm
. display 2*(asin(sqrt(0.5)) - asin(sqrt(0.25)))  // Cohen
.52359878

. display 2*(asin(sqrt(3/120)) - asin(sqrt(0)))   // Shaffer
.31756043

. display 2*(asin(sqrt(4/55)) - asin(sqrt(0)))    // Renaud
.54612129

Caution: Do not take the small, medium and large labels too seriously.

Q. What is the context for your question? Why do you want Cohen's h specifically? Why not a risk ratio or risk difference, for example?

HTH.

[Robert Murphy]

Hi Bruce, thank you very much for your helpful reply. I will try that.
The context is a choice experiment in which participants are provided with a description of 3 health states, told to imagine they would live in a health state and asked to select their 1st, 2nd and 3rd preference. In addition, the treatment group is provided, before making their choice, with information on the association between the health states and life satisfaction.
I test for differences between the control and treatment in the proportion selecting a health state as least preferred, 3rd preference, using prtesti
I plan to us Cohen's h to obtain a measure if effect size.

[Bruce Weaver]

Thanks for explaining the context, Robert. Has someone (e.g., a supervisor or boss) asked specifically for Cohen's h? Even if they have, I would also report the risk difference with a 95% CI. The risk difference is a much more intuitive measure of effect size, IMO.

net describe rdci, from(http://fmwww.bc.edu/RePEc/bocode/r)

HTH.