Pilot study power calculation (already read help menus and discussion board)

[Melissa Hagan]

Hello Everyone: Apologies if this has been discussed before and I did not locate it. Brief context: I'm conducting a pilot study but  the funders want power (and frankly so do I - I feel like I will retired before I am able to conduct a study that is not under-powered). I want to detect a difference of 1 (on a scale from 1 to 7) between two groups. Each group has 15 members (assuming that '15' goes into the n[within] field in the One-Way ANOVA applet?). I know which group will be higher than the other (i.e. I can predict direction). First question - I had used Cohen's as a guideline in the past but is there a better way to estimate the effect size if I have a specific value difference in mind that would be meaningful to me (such as the one above)?  Second, I don't know of another study done that can give me a sense of the other parameters needed (MSE) but (and apologies if this is a ridiculous thought) this measure (the one with a scale of 1-7) was used in a similar sample of 40 people and had a SD of 1.2 but not sure that is helpful here. Thank you so much! And thank you to this great resource - I've learned a lot from these pages and from the software already.

[Lenth, Russell V]

If you have just two groups, I think things will be less confusing if you use the app for the 2-sample t test, which is equivalent to ANOVA when you have two treatments. If you are confident that that SD of 1.2 represents variability of the same measure on similar people, it seems usable here. Put it for value of s1=s2, and 1.0 for the target difference between the means. If you are absolutely sure you want to do a one-sided test, check that box, but remember you can't really change your mind later as to the direction in which the groups will differ. Vary n1=n2 to obtain the required power.

Russ

Sent from my iPad

On Aug 13, 2014, at 1:57 PM, "Melissa Hagan" <meliss...@gmail.com> wrote:

Hello Everyone: Apologies if this has been discussed before and I did not locate it. Brief context: I'm conducting a pilot study but  the funders want power (and frankly so do I - I feel like I will retired before I am able to conduct a study that is not under-powered). I want to detect a difference of 1 (on a scale from 1 to 7) between two groups. Each group has 15 members (assuming that '15' goes into the n[within] field in the One-Way ANOVA applet?). I know which group will be higher than the other (i.e. I can predict direction). First question - I had used Cohen's as a guideline in the past but is there a better way to estimate the effect size if I have a specific value difference in mind that would be meaningful to me (such as the one above)?  Second, I don't know of another study done that can give me a sense of the other parameters needed (MSE) but (and apologies if this is a ridiculous thought) this measure (the one with a scale of 1-7) was used in a similar sample of 40 people and had a SD of 1.2 but not sure that is helpful here. Thank you so much! And thank you to this great resource - I've learned a lot from these pages and from the software already.

--
You received this message because you are subscribed to the Google Groups "PiFace discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to piface-discuss...@googlegroups.com.
To post to this group, send email to piface-d...@googlegroups.com.
Visit this group at http://groups.google.com/group/piface-discussion.
To view this discussion on the web visit https://groups.google.com/d/msgid/piface-discussion/568390e9-58d4-417c-9abe-dfd48030b5a9%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.