[R] Effect sizes

[Thomas Lumley]

t.test() does give effect sizes: that is, it gives the actual difference 
in means, and a confidence interval for this difference.

If you want to standardize your effect sizes you will have to do this 
yourself.  This is because the customs for standardization are different 
in different fields.  Many social sciences use the residual standard 
deviation, but epidemiology and medicine typically use the measured units 
(and where standardization is done, prefer to use %), and other 
standardizations are possible.

It's quite possible that there are packages that do all of this for you. 
If so, someone is likely to point them out.

 	-thomas


On Fri, 26 Oct 2007, Jenifer Larson-Hall wrote:

> I'm just curious . . . if effect sizes are so important, and possibly a 
> better way of looking at results than p-values, since they don't depend 
> on effect size (Kline,2004; Murphy and Myors, 2004), why don't any of 
> the classical tests, like t.test or glht specified for Tukey's posthocs, 
> return effect sizes? I say "classical" because I'm sure there may be 
> packages out there, not in the base program, which do return effect 
> sizes, but do they also return everything glht does, which are 
> confidence intervals for mean differences, t-values and p-values, 
> standard error plus a cool MMC graph? Anyway, just wondering. I mean, 
> it's not that hard to calculate effect sizes on my own, but it seems 
> like if they were important they would be included . . .
>
> Jenifer
>
> Dr. Jenifer Larson-Hall
> Assistant Professor of Linguistics
> University of North Texas
> (940)369-8950
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle