Dear all,
Is there anybody with experience of Edwards test for
seasonality with a variable population at risk as described
by Walter and Elwoods. I refer particularly to the estimate
of the amplitude of a simple harmonic trend. I am
constantly overestimating this and wonder whether my
interpretation of the estimate is wrong.
As far as i know, the estimate for alpha is:
Alpha= 2*[d*sqrt(k*M)-sum[sqrt(mi)*cos(thetai-thetastar)]
all divided by
sum[sqrt(mi)*cos(thetai-thetastar)^2]
where d is the distance of the sample centre of gravity
from zero.
k is the number of sectors (in my case twelve)
M is the total population at risk
mi is the population at risk per month
thetai is the monthly midangle
and
thetastar is the estimate of the direction for the maximum
rate
In short, is this correct (for instance where does the '2'
come from; my expansion abilities are not very good). If
not, any suggestions why my estimates of the amplitude are
so incorrect. (and thus i cannot calculate the goodness of
fit statistic, which is what i really want).
I am using excel to calculate it and it has been checked
thoroughly to see if it is holding the right formulae.
Any answers fully appreciated
Thanks
Paul
-----------------------------------
Paul Wicks
[log in to unmask]
St. Georges Hospital Medical School
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|