You can find my original query at the bottom of this mail. I got only 2 answers
on it, although some people asked for the summary, what indicates that this is
not a well known field.
The first response was
"This is a strange thing you want to do. However, i suggest that you treat it
as
a product moment correlation and use Fisher's z transformation, which gives an
approximately Normal variable with a standard error. You can then use this in
sample size calculations in the usual way." ([log in to unmask])
The other response was
"if a pilot sample is available you can try using the bootstrap technique
described in Modarres, Statistics and Computing, 1996, 6, pp. 139-145.
Otherwise a closed formula is available in Kraemer, 1974 JASA, 67, pp.114-117."
([log in to unmask])
The correct reference for Kraemer is "The non-null distribution of the
Spearman Rank Correlation Coefficient", H.C.Kraemer, March 1974, Vol.69,
N.345, pp.114-117. The other references I didn't check.
Many thanks to those who replied and those who gave my query a thought.
Kind regards,
Annelies De Brauwer
Biostatistician
Innogenetics, NV
Belgium, Gent
e-mail: [log in to unmask]
My original query:
"Dear AllStat'ers,
I need to calculate the sample size (N) for a case where I want to confirm that
there is a significant difference between a Spearman Rank correlation and a
hypothesized value. It would be appreciated if we could be pointed in the right
direction of some necessary literature (references), and/or the calculation
method, so that we can make some calculations for different alpha and beta
values.
Many thanks in advance.
Annelies"
|