Re: Sample size to study correlation [1 Attachment]

[Guest guest]

Hi Smita,I presume you used Fisher's z-transformation to normalize the distribution of correlation coefficient (in that case, should this not apply only to Pearson's correlation? What happens with Spearman's or Kendall's rank correlation?)

This is the z transformation that z^' = 0.5 * ln([1+r] / [1-r]) where z^' = transformed z, r = correlation coefficient for the corresponding z, ie r0, and r1, ln = natural logarithmYou can get the corresponding z from z^' therefore by the following formula:

z = z^' - ((3 * z^' + r) / 4 * n)These formulae were taken from the Statsdirect website:http://www.statsdirect.com/help/sample_size/sscorr.htm

Best,Arin

On Sun, Jul 11, 2010 at 2:30 AM, smita mali <smt_mali@...> wrote:

[Attachment(s) from smita mali included below]

Hello Sir,

For calculating the sample size to do the correlation studies here is the formula

N= [(z1-á/2 + z1-â) / |z’r0 – z’r1|]2 + 3

where,

Information required

POWER: probability of detecting a true effect

ALPHA: probability of detecting a false effect (two sided: double this if you need one sided)

R0: correlation coefficient under the null hypothesis (often 0)

R1: correlation coefficient under the alternative hypothesis

I am attaching the reference document and explanation for the derivation of this formula.

Sir, I could not found anything simpler than this. It will be definitely welcome to hear from NetRUMians if anyone would come up with more information.

Regards,

Dr Smita Mali