[R] Power calculation for detecting linear trend

[Peter Dalgaard]

Meesters, Erik wrote:
> Dear people,
> I've a problem in doing a power calculation. In Fryer and Nicholson
> (1993), ICES J. mar. Sci. 50: 161-168 page 164 an example is given with
> the following characteristics
> T=5, points in time
> R=5, replicates
> Var.within=0.1
> q=10, a 10% increase per year
> The degrees of freedom for the test are calculated as Vl=T*R-2=23 and
> the non-centrality parameter Dl=4.54.
> Using this they get a power of 0.53, but the result that I'm getting is
> 0.05472242.
>
> I've tried this several ways in R, but I'm not able to come up with the
> same number. Am I doing something wrong in the calculation of the power?
> Here's my code:
>
>     T<-5
>     R<-5
>     sigmasq<-0.1
>     q<-10
>     Vl<-(T*R)-2
>     Dl<-(R*(T-1)*T*(T+1)/(12*sigmasq))*(log(1+(q/100)))^2 #Dl result is
> still similar
>
>     power.1<-1-pf(qf(.95,(T*R-2),1,ncp=0),(T*R-2),1,ncp=Dl)
>
> Thank you for any suggestions/help.
>   
I think your DF are upside-down:

> power.1<-1-pf(qf(.95,1,(T*R-2),ncp=0),1,(T*R-2),ncp=Dl)
> power.1
[1] 0.532651



-- 
   O__  ---- Peter Dalgaard             Øster Farimagsgade 5, Entr.B
  c/ /'_ --- Dept. of Biostatistics     PO Box 2099, 1014 Cph. K
 (*) \(*) -- University of Copenhagen   Denmark          Ph:  (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)                  FAX: (+45) 35327907